Showing 1–50 of 971 results for author: He, B

AI4AI-Bench: Benchmarking LLM Agents in Algorithmic Design for Recursive Self-Improvement

Authors: Yizhe Chi, Wenyi Li, Deyao Hong, Xiaoqiu Wang, Mingju Gao, Kaisen Yang, Bingxiang He, Youjie Zheng, Calvin Xiao, Qinhuai Na

Abstract: Recursive self-improvement (RSI) asks whether an AI system can improve the process that produces AI systems, so that the next system inherits the improvement. That process is the training algorithm: a better objective or update rule improves the compute\mbox{-}capability exchange rate for every subsequent run, including the one that produces the next agent. Whether RSI is feasible therefore turns… ▽ More Recursive self-improvement (RSI) asks whether an AI system can improve the process that produces AI systems, so that the next system inherits the improvement. That process is the training algorithm: a better objective or update rule improves the compute\mbox{-}capability exchange rate for every subsequent run, including the one that produces the next agent. Whether RSI is feasible therefore turns on whether an agent can design training algorithms. No benchmark isolates that ability: existing suites are won by collecting data or by tuning hyperparameters, and none tells a change to how a run is executed apart from a change to how the model learns. We present AI4AI\mbox{-}Bench, 10 frozen research repositories spanning 10 training algorithm families. In each task, an agent has 4 hours on one B300 to rewrite the training algorithm; its code is then rerun from scratch for up to 12 hours and scored by a fixed evaluator hidden from the agent, against the repository's original algorithm under the same procedure. Because the 10 metrics are incommensurable, every task is mapped onto one scale on which $0$ is an uninformative model, $0.1$ is the algorithm the repository ships, and $1.0$ is the task optimum. Across 29 configurations of 6 systems on all 10 tasks the mean score is $0.166$, and the best system reaches $0.250$: even the strongest closes under a fifth of the distance between the algorithm that was already there and the optimum. The submissions show where that distance went: most never change how the model learns at all, and the minority that do average $0.226$ against $0.126$ for the rest. More reasoning effort mostly buys the willingness to go there, taking that minority from $8\%$ of submissions to $64\%$ and the mean score from $0.094$ to $0.196$. We release the task suite, the evaluators and every scored submission, so that the measurement can be repeated as these systems change. △ Less

Submitted 20 August, 2026; originally announced August 2026.

Open-MOPD: Diagnosing and Fixing Capability Imbalance in Multi-Teacher On-Policy Distillation

Authors: Huan-ang Gao, Haohan Chi, Yong Yan, Shiyuan Feng, Hanlin Wu, Zheng Jiang, Bingxiang He, Wei-Ying Ma, Ya-Qin Zhang, Hao Zhou

Abstract: Multi-teacher on-policy distillation (M-OPD) has emerged as a promising paradigm for consolidating domain-specialized reinforcement learning (RL) experts into a single generalist student via dense, token-level reward supervision. Despite its practical success, the optimization dynamics governing multi-teacher capability integration remain poorly understood, and open, rigorously reproducible recipe… ▽ More Multi-teacher on-policy distillation (M-OPD) has emerged as a promising paradigm for consolidating domain-specialized reinforcement learning (RL) experts into a single generalist student via dense, token-level reward supervision. Despite its practical success, the optimization dynamics governing multi-teacher capability integration remain poorly understood, and open, rigorously reproducible recipes are conspicuously lacking. In this work, we establish a controlled M-OPD benchmark on SmolLM3-3B-Base with oracle routing, isolating capability integration from routing ambiguity. Our investigation reveals a pronounced capability integration gap: standard M-OPD captures only 35.6% of the available headroom relative to a domain-routed oracle ensemble, with concise tasks such as instruction following suffering severe degradation and premature stagnation. Crucially, we show that this failure stems not from gradient conflict, but from a severe misallocation of the token-level optimization budget. This pathology is driven by three orthogonal factors: structural sequence-length disparities across domains, dynamic convergence drift due to non-uniform learning rates, and multi-step reward staleness from asynchronous policy updates. To resolve these imbalances, we introduce Open-MOPD, a principled framework incorporating token-share balancing, gap-aware dynamic budget allocation, and student reward refresh. Together, these mechanisms systematically restore cross-domain balance, elevating headroom recovery from 35.6% to 83.4% in a single deployable student. We fully open-source our end-to-end post-training recipe, training trajectories, and evaluation suites on an academically accessible hardware budget. △ Less

Submitted 19 August, 2026; originally announced August 2026.

Comments: Project page: https://bytedtsinghua-sia.github.io/Open-MOPD/

Evo-Harness: Context-to-Harness Skill Compilation for Self-Evolving Agents

Authors: Tianxin Wei, Zhan Shi, Minhua Lin, Bing He, Zewen Liu, Yisi Sang, Yuanchen Bei, Xuying Ning, Jiaru Zou, Ting-Wei Li, Xiao Lin, Yanjun Zhao, Chi Wang, Benoit Dumoulin, Dakuo Wang, Jingrui He, Hanqing Lu

Abstract: Learning from experience is critical for developing capable, self-improving large language model (LLM) agents. Existing methods typically extract knowledge from accumulated trajectories via reflection, memory, rules, or skills. However, agents in realistic environments continuously encounter novel tasks, often offering only a one-shot opportunity to improve. These executions yield rich but highly… ▽ More Learning from experience is critical for developing capable, self-improving large language model (LLM) agents. Existing methods typically extract knowledge from accumulated trajectories via reflection, memory, rules, or skills. However, agents in realistic environments continuously encounter novel tasks, often offering only a one-shot opportunity to improve. These executions yield rich but highly noisy contexts, entangling broadly useful lessons with task-specific artifacts. Critically, prior works rarely validate their effectiveness on complex real-world tasks or isolate the underlying drivers of improvement. To address these gaps, we formulate online harness learning, where a frozen agent improves by continually updating a structured harness across sequential tasks. This formulation enables a systematic study of key self-improvement factors through our proposed Evo-Harness. At its core, context-to-harness skill compilation distills noisy, single-shot executions into reusable skill harnesses for cross-domain and topic-level adaptation. To demonstrate the efficacy of one-shot skill compilation, we evaluate across five realistic benchmarks (TerminalBench2, SWE-bench, CL-Bench, -bench, WebArena-Infinity). Our extensive analysis demonstrates the effectiveness of Evo-Harness and provides a principled understanding of how LLM agents can effectively learn on the fly. Our code is available at https://github.com/A-EVO-Lab/a-evolve/tree/release/evo-harness. △ Less

Submitted 15 August, 2026; originally announced August 2026.

PACE-Bench: Benchmarking Physics Adaptation via Code Evolution in Dynamic Environments

Authors: Yuhao Zhan, Bingxiang He, Zecong Tang, Chaojun Xiao

Abstract: Self-evolving agents improve future behavior from interaction experience, yet existing evaluations typically optimize under fixed execution conditions and do not test recovery after those conditions change. To address this gap, we introduce PACE-Bench (Physics Adaptation via Code Evolution), a simulator-grounded benchmark of 144 source-to-target adaptation pairs across six physics domains. Each pa… ▽ More Self-evolving agents improve future behavior from interaction experience, yet existing evaluations typically optimize under fixed execution conditions and do not test recovery after those conditions change. To address this gap, we introduce PACE-Bench (Physics Adaptation via Code Evolution), a simulator-grounded benchmark of 144 source-to-target adaptation pairs across six physics domains. Each pair links a source environment to a mutated target environment with the same goal and interface. A code-driven design that succeeds in the source fails in the target, where agents must iteratively adapt it into a working target design using diagnostic sandbox feedback within a limited attempt budget. We compare ten self-evolving methods from four paradigms. The benchmark remains far from saturated: Reflexion + Qwen3-14B succeeds on only 35.9\% of full-benchmark pairs, while GPT-5.5 solves 66.7\% of the Statics subset under the full budget. Together, these results show that simulator-grounded reflection is more reliable than unverified self-revision, while memory anchors agents to early designs and broad tree search explores without converging. Even revealing exact physical changes does not raise the performance ceiling, pointing to mechanism redesign rather than parameter inference as the central bottleneck. Data and code are available at https://github.com/thunlp/PACE-Bench. △ Less

Submitted 14 August, 2026; originally announced August 2026.

MathForm: Scaling Mathematical Autoformalization with Knowledge Retrieval and Verification-Guided Refinement

Authors: Lushi Pu, Weiming Zhang, Xinheng Xie, Zixuan Fu, Bingxiang He, Hengyu Zhao, Hongya Lyu, Xin Li, Jie Zhou, Yudong Wang

Abstract: Autoformalization is commonly framed as translating natural-language mathematical statements into machine-verifiable formal languages such as Lean 4. However, faithful formalization requires more than translation. Models must map mathematical concepts to the complex hierarchy of types and definitions in formal libraries such as Mathlib, while ensuring that generated statements preserve the meaning… ▽ More Autoformalization is commonly framed as translating natural-language mathematical statements into machine-verifiable formal languages such as Lean 4. However, faithful formalization requires more than translation. Models must map mathematical concepts to the complex hierarchy of types and definitions in formal libraries such as Mathlib, while ensuring that generated statements preserve the meaning of the source propositions. Existing approaches struggle because they rely heavily on the model's parametric memory for library-specific knowledge, while common data construction pipelines often resort to filtering single-pass outputs and lack mechanisms for feedback-driven revision. To address these challenges, we introduce MathForm, an autoformalization framework for constructing verified training data through Mathlib knowledge retrieval and verification-guided iterative refinement. Before generation, a retrieval planner gathers relevant definitions and existing formalizations from Mathlib to guide the formalization generator. Generated statements are then revised using compiler diagnostics and semantic-consistency feedback. Using this framework, we construct FormalVerse, a Lean 4 dataset containing approximately 367K verified examples across diverse mathematical domains and sources. We then train MathForm-8B through supervised fine-tuning followed by reinforcement learning. Across six benchmarks, MathForm-8B achieves average Pass@8 rates of 88.06% under Syntax Check (SC) and 72.37% under Consistency Check (CC), outperforming multiple specialized 32B autoformalizers. On the challenging FATE-H and FATE-X subsets, it attains CC pass rates of 63% and 37%, exceeding the strongest specialized baselines in both cases. △ Less

Submitted 14 August, 2026; originally announced August 2026.

Comments: 25 pages, 6 figures, 8 tables

GESTO: Human-Centric Spatio-Temporal Memory for Reasoning in Dynamic Scenes

Authors: Ermanno Bartoli, Buwei He, Dennis Rotondi, Sebastian Koch, Federico Tombari, Kai O. Arras, Patric Jensfelt, Yixi Cai, Iolanda Leite

Abstract: Robots operating in human environments need memories that capture not only what objects exist and where, but also how people use them over time and how individual interactions compose into goal-directed activities. Existing 4D scene graphs preserve object and place histories but omit activity structure, whereas activity representations are either not grounded in persistent 3D scenes or rely on ext… ▽ More Robots operating in human environments need memories that capture not only what objects exist and where, but also how people use them over time and how individual interactions compose into goal-directed activities. Existing 4D scene graphs preserve object and place histories but omit activity structure, whereas activity representations are either not grounded in persistent 3D scenes or rely on externally provided event boundaries and object associations. We present GESTO (Grounded Event and Spatio-Temporal memOry), a spatio-temporal memory that couples a persistent 4D scene graph with a two-level hierarchy of atomic human--object interactions and goal-driven events. From an RGB-D observation stream, GESTO automatically extracts timestamped interactions, grounds them to persistent scene entities, groups them into events, and uses event context to refine uncertain object associations. A relation-aware tool-calling agent queries the resulting memory for activity-centric spatio-temporal reasoning. We evaluate GESTO on the reproducible text, binary, and time categories of an existing benchmark, together with 40 new Space2Event and Event2Space queries. GESTO achieves scores of 0.71, 0.75, and 0.70 on the standard categories, approaching a method supplied with ground-truth event and object grounding, while substantially outperforming the same reasoning framework when these inputs are removed. It further achieves 0.73 and 0.75 on Space2Event and Event2Space queries. Ablations show that hierarchical event structure and context-aware grounding refinement provide complementary benefits, supporting activity-grounded hierarchical memory for retrospective reasoning in dynamic human environments. △ Less

Submitted 11 August, 2026; originally announced August 2026.

MIRA: Medical Image Reflection for Agentic Diagnosis

Authors: Shengzhi Wang, Jun Yang, Kai Wu, Xiaozhong Ji, Yiwen Ye, Ziyang Chen, Mingliang Xiong, Wen Fang, Mingqing Liu, Mengyuan Xu, Miaoxuan Shan, Caiyan Liu, Bin He, Qingwen Liu

Abstract: Medical visual agents can use tools to inspect images and retrieve external knowledge, but indiscriminate tool use may introduce noisy or misleading evidence. Reliable diagnosis therefore requires not only acquiring additional observations, but also verifying whether tool actions are necessary and whether the resulting evidence supports the current hypothesis. We introduce MIRA (Medical Image Refl… ▽ More Medical visual agents can use tools to inspect images and retrieve external knowledge, but indiscriminate tool use may introduce noisy or misleading evidence. Reliable diagnosis therefore requires not only acquiring additional observations, but also verifying whether tool actions are necessary and whether the resulting evidence supports the current hypothesis. We introduce MIRA (Medical Image Reflection for Agentic Diagnosis), a medical visual diagnostic framework for autonomous evidence search and reflective verification. MIRA dynamically invokes image-processing operations, including zooming, grounding, pointing, rotation, and measurement, as well as web search, while evaluating the relevance and consistency of the acquired evidence. We develop MIRA through a two-stage training strategy. First, a tool-augmented Monte Carlo Tree Search data engine explores diverse diagnostic hypotheses and jointly verifies visual grounding accuracy and semantic consistency to construct supervised fine-tuning trajectories. Second, reinforcement learning further improves decision-making through online reflective principle evolution: failure cases are distilled into candidate principles, and only principles that improve held-out rollout rewards are retained. Across nine medical visual reasoning benchmarks, MIRA achieves an average score of 64.73, improving its Qwen3-VL-8B backbone by 7.44 points. It also increases useful tool-use judgments from 56.2% to 73.8% and reduces harmful judgments from 8.9% to 1.6%. Qualitative analyses show that MIRA can re-examine evidence, correct premature conclusions, and adapt its tool-use strategy. Project page: https://MIRA-VL.github.io/ △ Less

Submitted 11 August, 2026; originally announced August 2026.

DocTrace: Towards Traceable Long Document VQA via Hierarchical Evidence Graph Reasoning

Authors: Le Xiang, Zhicheng Guan, Hong Chen, Xiaocong Lin, Zhenghua Lei, Teng Hu, Bolei He, Long Zeng

Abstract: Long Document Visual Question Answering (LongDocVQA) requires Multimodal Large Language Models (MLLMs) to locate, integrate, and reason over heterogeneous document elements distributed across multiple pages. Existing approaches, including end-to-end MLLMs, retrieval-augmented generation (RAG) pipelines, and document agents, often lack explicit mechanisms to represent and verify how grounded eviden… ▽ More Long Document Visual Question Answering (LongDocVQA) requires Multimodal Large Language Models (MLLMs) to locate, integrate, and reason over heterogeneous document elements distributed across multiple pages. Existing approaches, including end-to-end MLLMs, retrieval-augmented generation (RAG) pipelines, and document agents, often lack explicit mechanisms to represent and verify how grounded evidence is progressively composed during reasoning, limiting both answer accuracy and traceability. In this paper, we cast LongDocVQA as an explicit evidence graph reasoning problem rather than implicit answer prediction. To this end, we propose DocTrace, a hierarchical framework that progressively performs evidence localization, structured document parsing, and evidence graph reasoning to enable explicit evidence provenance. To effectively learn these capabilities, we develop a two-stage training framework: joint Supervised Fine-Tuning (SFT) first initializes evidence localization and graph reasoning abilities, followed by task-specific Group Relative Policy Optimization (GRPO) with dedicated rewards to further optimize these capabilities. Extensive experiments on MMLongBench-Doc, LongDocURL, and SlideVQA demonstrate that DocTrace consistently outperforms both existing open-source baselines and proprietary MLLMs. Compared with the Qwen3-VL-8B-Instruct backbone, DocTrace achieves absolute improvements of 14.4, 11.3, and 11.7 points on the three benchmarks, respectively. Beyond competitive performance, DocTrace constructs traceable evidence graphs with explicit node-level provenance, enabling transparent and verifiable reasoning for long document understanding. △ Less

Submitted 4 August, 2026; originally announced August 2026.

TRACE: Ergodic Trajectory Optimization for Active Scene Reconstruction

Authors: Ziyue Zheng, Linli Shi, Bingkun He, Wen Jiang, Ziyun Wang

Abstract: Existing active reconstruction systems with Gaussian-splatting maps select observations greedily, optimizing a single next-best-view (NBV) at each step and connecting the chosen views by short-horizon path planning. This greedy decoupling disregards the global structure of scene information, producing inefficient trajectories that waste sensing capacity in transit between selected views. In this w… ▽ More Existing active reconstruction systems with Gaussian-splatting maps select observations greedily, optimizing a single next-best-view (NBV) at each step and connecting the chosen views by short-horizon path planning. This greedy decoupling disregards the global structure of scene information, producing inefficient trajectories that waste sensing capacity in transit between selected views. In this work, we study active reconstruction as an ergodic coverage problem: the time-averaged spatial statistics of the sensor trajectory should match a target information distribution induced by the current map. Our approach derives this target distribution online from uncertainty and visibility, and calculates ergodic trajectories via a kernel-ergodic horizon planner with gradient flow and footprint depletion, closing the loop between mapping and trajectory optimization. We thoroughly evaluate TRACE on the Replica dataset against the Next-Best-View (NBV) baselines, improving PSNR by 1.5 dB. Code: https://github.com/spikelab-jhu/trace-active-reconstruction. △ Less

Submitted 5 August, 2026; v1 submitted 3 August, 2026; originally announced August 2026.

Comments: 11 pages, 7 figures, fixed a template bug in the Latex

AdaMTP: An Adaptive Training Paradigm for Multi-Token Prediction

Authors: Ziqiang Cui, Han Shi, Bowei He, Yu Pan, Peiyang Liu, Shengyin Sun, Yankai Chen, Haoli Bai, Yichun Yin, Xue Liu, Chen Ma

Abstract: Multi-Token Prediction (MTP) has emerged as an effective paradigm that augments a shared Large Language Model backbone with auxiliary heads, training the model to predict several future tokens in parallel to enrich its supervision signal and accelerate inference. However, existing training frameworks adopt a rigid, fixed-length prediction horizon, disregarding the highly non-uniform information de… ▽ More Multi-Token Prediction (MTP) has emerged as an effective paradigm that augments a shared Large Language Model backbone with auxiliary heads, training the model to predict several future tokens in parallel to enrich its supervision signal and accelerate inference. However, existing training frameworks adopt a rigid, fixed-length prediction horizon, disregarding the highly non-uniform information density of natural language and code. Forcing the auxiliary heads to predict across high-entropy semantic boundaries injects noisy, conflicting training signals; because these heads share the backbone's latent representations, the resulting gradients backpropagate and interfere with the model's core capabilities. We propose AdaMTP, an adaptive training paradigm that dynamically aligns the prediction horizon with the intrinsic predictability of the sequence. At its core, an entropy-based segmentation algorithm leverages the base model to detect sudden surges in uncertainty as semantic boundaries, partitioning sequences into variable-length groups. Each token is assigned an adaptive prediction depth, and a dynamically masked MTP objective suppresses the loss for predictions that cross these boundaries, attenuating the noisy gradients that degrade the backbone. Across mathematical reasoning, code generation, and general benchmarks on three backbones (Llama-3.1-8B, Qwen-2.5-7B, Gemma-3-12B), AdaMTP consistently outperforms standard MTP in both task performance and inference speedup. △ Less

Submitted 1 August, 2026; originally announced August 2026.

Chain-of-Models: Cross-Model Auditing for Bias-Robust LLM Judges

Authors: Qian Wang, Zhanzhi Lou, Zhenheng Tang, Nuo Chen, Bingsheng He

Abstract: LLMs increasingly serve as automated judges, but their judgments remain vulnerable to cognitive biases. Existing mitigations mostly rely on prompt-driven debiasing, which is brittle across bias types, or human evaluation, which does not scale. We study \emph{Chain-of-Models} (CoM), an automated audit pipeline in which a second model inspects the first model's reasoning trace before producing the f… ▽ More LLMs increasingly serve as automated judges, but their judgments remain vulnerable to cognitive biases. Existing mitigations mostly rely on prompt-driven debiasing, which is brittle across bias types, or human evaluation, which does not scale. We study \emph{Chain-of-Models} (CoM), an automated audit pipeline in which a second model inspects the first model's reasoning trace before producing the final judgment. The key design question is whether the auditor should be the same model, a same-family model, or a different-family model. Across 9 models from 6 families, 4 cognitive biases, and 4 factual datasets, we find that auditor identity matters in two ways. First, standalone bias resistance does not predict audit effectiveness: Kimi-K2.5 is the strongest standalone model on several biases, yet is a weak auditor for Qwen2.5-72B's biased traces. Second, the best auditor is bias-specific: GPT-4o is strongest on bandwagon, authority, and distraction, while GLM-5 is strongest on sycophancy. We operationalize these findings with a per-bias auditor selection rule that, given the bias type, scores candidates along functional diversity, per-bias standalone resistance, and calibrated audit effectiveness. Under a calibration/test split, the selector reaches the highest accuracy across the four biased slices ($0.884$ vs.\ $0.824$ for the strongest single fixed auditor and $0.805$ for the no-audit baseline). We release data, configurations, and an LLM-agent skill at https://anonymous.4open.science/r/chain-of-models-B585 . △ Less

Submitted 19 May, 2026; originally announced July 2026.

Comments: WIP

Restoring Collaborative Signals in Semantic-ID Generative Recommendation via Personalized Natural Language

Authors: Changjiang Han, Qingyang Li, Yaqiang Zang, Jikun Kang, Pinghua Gong, Xue Liu, Bowei He

Abstract: Making LLM-based generative recommendation models stronger and more personalized through natural language and explicit reasoning is a widely anticipated yet still unsolved goal. Such models cast recommendation as autoregressively generating an item's semantic-ID (SID), a short tuple of discrete codes, so that recommending well reduces to emitting the right SID. In this setting the model verbalizes… ▽ More Making LLM-based generative recommendation models stronger and more personalized through natural language and explicit reasoning is a widely anticipated yet still unsolved goal. Such models cast recommendation as autoregressively generating an item's semantic-ID (SID), a short tuple of discrete codes, so that recommending well reduces to emitting the right SID. In this setting the model verbalizes its knowledge poorly, and text and SID tokens live in misaligned embedding spaces. Deep reasoning therefore rarely turns into a correct SID, and enabling explicit "thinking" often gives no gain or even hurts. The deeper cause is that a compact SID cannot hold content and collaborative signal at once: the two compete, and collaboration loses. Because a mis-predicted SID is a wrong recommendation, this caps accuracy directly. Costly multi-round training barely helps, and few methods try to supply the missing signal at inference time. What is missing is a reliable channel that carries collaborative signal into SID generation. We therefore propose a framework, guided by personalized natural language, that adds hierarchical collaborative cues as the model generates, without altering the backbone or retraining the SIDs. Rather than mapping language onto SIDs directly, it uses language to attach analyzable links between collaborative patterns and their audiences, restoring the collaborative signal that SIDs miss. The result is consistent gains in recommendation accuracy, grounding generation in collaborative structure at inference time rather than relying on explicit reasoning or retraining. △ Less

Submitted 30 July, 2026; originally announced July 2026.

Comments: 8 pages, 4 figures

AE-UAV: An Air-to-Air Event-Based UAV Tracking Benchmark and a Real-Time Frequency-Domain Tracker

Authors: Zixin Jiang, Bing He, Chaoran Xiong, Zhenzhen Wang, Xin Zhao, Ling Pei

Abstract: Air-to-air (A2A) unmanned aerial vehicle (UAV) tracking is fundamental to airborne remote sensing of low-altitude aerial targets. However, the deployment of continuous, real-time tracking systems on UAVs presents significant challenges. In A2A scenarios, traditional frame-based cameras suffer from severe performance degradation under low illumination, overexposure, and high-speed motion owing to t… ▽ More Air-to-air (A2A) unmanned aerial vehicle (UAV) tracking is fundamental to airborne remote sensing of low-altitude aerial targets. However, the deployment of continuous, real-time tracking systems on UAVs presents significant challenges. In A2A scenarios, traditional frame-based cameras suffer from severe performance degradation under low illumination, overexposure, and high-speed motion owing to their limited dynamic range and fixed temporal sampling. Although event cameras offer a promising alternative with microsecond temporal resolution and a high dynamic range, current research is bottlenecked by two primary issues: 1) the absence of dedicated A2A event-based datasets, and 2) the heavy reliance of existing trackers on GPU acceleration and extensive training data, rendering them impractical for resource-constrained UAVs. To bridge these gaps, we introduce AE-UAV, an air-to-air event-based UAV tracking benchmark. To the best of our knowledge, this is the first airborne-captured event camera dataset for A2A tracking, comprising 178 flight sequences with continuous-time cubic B-spline annotations. Furthermore, we propose the Fast-Slow Frequency-domain Tracking (FSFT) method. This lightweight, training-free framework seamlessly integrates frequency-domain template matching with search region prediction and detection-based drift correction. Extensive experiments demonstrate that FSFT operates at an ultra-fast 420 frames per second (FPS) on CPU-only hardware. It retains 93.97% of the accuracy of state-of-the-art GPU-dependent methods while delivering a 5.32-fold effective speedup and exhibiting superior temporal resolution generalization, thereby providing a highly efficient and robust solution for airborne remote sensing of aerial targets. The dataset and source code are available at https://github.com/MSP-xEN/AE-UAV. △ Less

Submitted 16 July, 2026; originally announced July 2026.

Comments: 12 pages, 7 figures. Submitted to IEEE Transactions on Geoscience and Remote Sensing

Representation-Aligned Tactile Grounding for Contact-Rich Robotic Manipulation

Authors: Ruilin Chen, Jingkai Jia, Tong Yang, Xinyu Zhou, Qiao Sun, Jiangwei Zhong, Shizeng Zhang, Nuo Chen, Bailin He, Wei Li, Wenqiang Zhang

Abstract: Tactile-enhanced vision-language-action (VLA) policies have been introduced for contact-rich manipulation, where critical interaction states are often hidden from vision. Future tactile prediction is a promising way to use touch because it turns tactile outcomes into supervision for action-induced contact dynamics. Yet VLA policies contain representations with different roles, from perceptual enco… ▽ More Tactile-enhanced vision-language-action (VLA) policies have been introduced for contact-rich manipulation, where critical interaction states are often hidden from vision. Future tactile prediction is a promising way to use touch because it turns tactile outcomes into supervision for action-induced contact dynamics. Yet VLA policies contain representations with different roles, from perceptual encoding to motor prediction, making it unclear where this supervision should be applied. We study this as a representation-alignment problem. Through a linear probe analysis, we find that future tactile states are most predictable from intermediate action-expert features, rather than from vision-language features or final action states. Motivated by this observation, we introduce a lightweight Latent Tactile Predictor (LTP), which predicts compact future tactile embeddings from the identified intermediate representation. By avoiding direct prediction of noisy raw tactile signals, LTP provides an action-outcome grounding signal that aligns intermediate action representations with future contact consequences. Experiments on real-world contact-rich manipulation tasks show that representation-aligned tactile grounding outperforms less aligned or multi-interface tactile prediction, highlighting the importance of where tactile supervision is applied. △ Less

Submitted 16 July, 2026; originally announced July 2026.

Branching Policy Optimization: Sandbox-Native Language Agent Reinforcement Learning

Authors: Bowei He, Yankai Chen, Xiaokun Zhang, Xue Liu

Abstract: Reinforcement learning has emerged as the dominant paradigm for training large language model (LLM) agents that interact with executable sandboxes. State-of-the-art algorithms such as PPO, RLOO, and GRPO inherit their rollout topology from RLHF: for each prompt, N independent trajectories are sampled from the initial state, and an advantage is computed by subtracting a group baseline. This design… ▽ More Reinforcement learning has emerged as the dominant paradigm for training large language model (LLM) agents that interact with executable sandboxes. State-of-the-art algorithms such as PPO, RLOO, and GRPO inherit their rollout topology from RLHF: for each prompt, N independent trajectories are sampled from the initial state, and an advantage is computed by subtracting a group baseline. This design ignores a defining property of agent sandboxes. They are deterministic, snapshottable, and resumable from any intermediate state. We argue that this property enables a fundamentally different rollout topology: rather than N independent trees of depth T, one can construct a single tree of N leaves whose siblings share prefixes, and therefore share variance. We instantiate this idea as Branching Policy Optimization (BPO), a sandbox-native RL algorithm that (i) adaptively snapshots the sandbox at high-entropy decision points along a backbone trajectory, (ii) forks K alternative actions per branch point and rolls out each to termination, and (iii) computes per-step advantages from sibling returns rather than from independent prompts. We prove this estimator is unbiased and has strictly lower variance than the trajectory-level baseline, with the reduction equal to the prefix-explained portion of return variance. On WebShop, ALFWorld, and SWE-bench Verified with Qwen2.5-7B and Llama-3.1-8B backbones, BPO improves success by 3.6--6.1 absolute points over GRPO and RLOO at matched compute, halves gradient-norm variance, and matches the best baseline using 38% fewer policy updates. △ Less

Submitted 15 July, 2026; originally announced July 2026.

Comments: Accepted by WAIC Academic 2026

NodeImport: Imbalanced Node Classification with Node Importance Assessment

Authors: Nan Chen, Zemin Liu, Bryan Hooi, Bingsheng He, Jun Hu, Jia Chen

Abstract: In real-world applications, node classification on graphs often faces the challenge of class imbalance, where majority classes dominate training, resulting in biased model performance. Traditional GNNs often struggle in such scenarios, as they tend to overfit to majority classes while underrepresenting minority classes. Existing solutions, which either prioritize nodes based on class size or synth… ▽ More In real-world applications, node classification on graphs often faces the challenge of class imbalance, where majority classes dominate training, resulting in biased model performance. Traditional GNNs often struggle in such scenarios, as they tend to overfit to majority classes while underrepresenting minority classes. Existing solutions, which either prioritize nodes based on class size or synthesize new nodes for minority classes, often fall short of effectively addressing this imbalance issue. This paper introduces an approach to class-imbalanced node classification by utilizing a balanced meta-set for importance measurement, where a training node is considered significant if it enhances model performance under an unbiased setting. Our method identifies important nodes that can counteract class imbalance and utilizes them for model training, allowing for fine-grained and dynamic node selection throughout the training process. We theoretically derive a formula to directly assess node importance, reducing computational overhead and providing an intuitive threshold for node selection. Guided by this metric, we develop a novel framework that filters valuable labeled, unlabeled, and synthetic nodes that enhance model performance in an unbiased context. A key advantage of this framework is its separation of the synthetic node generation process from the filtering process, ensuring compatibility with various node generation methods. Furthermore, we introduce a strategy to construct a high-quality meta-set that closely approximates the overall feature distribution, ensuring robust representation of each class. We evaluate our framework, NodeImport, across multiple datasets using popular GNN architectures, demonstrating its superiority over existing baselines. Our results highlight the flexibility and effectiveness of the framework in mitigating class imbalance, leading to improved outcomes. △ Less

Submitted 15 July, 2026; originally announced July 2026.

Journal ref: KDD '25: Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.1, 2025, Pages 94 - 105

Discrete Diffusion Models: A Unified Framework from Tokenization to Generation

Authors: Ye Yuan, Weien Li, Rui Song, Zeyu Li, Haochen Liu, Xiangyu Kong, Zixuan Dong, Linfeng Du, Zipeng Sun, Weixu Zhang, Jiaxin Huang, Changjiang Han, Yonghan Yang, Zichen Zhao, Xiuyuan Hu, Haolun Wu, Yankai Chen, Fengran Mo, Jikun Kang, Bowei He, Philip S. Yu, Xue Liu

Abstract: Discrete denoising diffusion models (DDMs) have recently emerged as a compelling alternative to autoregressive (AR) modeling for discrete data, offering parallel generation and iterative global refinement capabilities. Unlike continuous diffusion, where the state space is fixed, DDMs are fundamentally shaped by how the discrete state space is constructed: the tokenization scheme, the vocabulary to… ▽ More Discrete denoising diffusion models (DDMs) have recently emerged as a compelling alternative to autoregressive (AR) modeling for discrete data, offering parallel generation and iterative global refinement capabilities. Unlike continuous diffusion, where the state space is fixed, DDMs are fundamentally shaped by how the discrete state space is constructed: the tokenization scheme, the vocabulary topology, and domain-specific structural alphabets. This work introduces a unified conceptual framework that views discrete diffusion models through the construction of the underlying discrete state space. Within this framework, existing formulations, including transition-matrix, masking/absorbing-state, and score/ratio-based approaches, emerge as different instantiations of a common design space. The framework further exposes common design trade-offs across training objectives, inference algorithms, scaling behavior, systems optimization, and evaluation protocols, suggesting several promising directions for future research. △ Less

Submitted 15 July, 2026; originally announced July 2026.

Compound Interference Recognition for LR-FHSS Satellite IoT Uplinks via Multi-Domain Instance Fusion

Authors: H. Xu, B. He, S. Wang, Y. Jiang

Abstract: Long range-frequency hopping spread spectrum (LR-FHSS) is a promising uplink physical layer for massive low Earth orbit satellite Internet of Things, where low power terminals report short packets from wide area regions with limited terrestrial infrastructure. However, satellite IoT links are exposed to external interference, and the coexistence of multiple interference components can severely deg… ▽ More Long range-frequency hopping spread spectrum (LR-FHSS) is a promising uplink physical layer for massive low Earth orbit satellite Internet of Things, where low power terminals report short packets from wide area regions with limited terrestrial infrastructure. However, satellite IoT links are exposed to external interference, and the coexistence of multiple interference components can severely degrade receiver reliability and complicate interference mitigation. Existing recognition methods either focus on single interference scenarios or treat each compound interference combination as an independent class, leading to limited generalization or poor scalability. To address this problem, this paper formulates LR-FHSS uplink compound interference recognition as a multi-instance multi-label learning problem and proposes a multi-domain instance fusion method. The proposed method fuses local instances from the time-frequency and frequency domains and aggregates their predictions for bag-level multi-label recognition. A dataset construction pipeline is developed based on the US915 LR-FHSS configuration and incorporates shadowed-Rician fading and time-varying Doppler to emulate practical satellite communication conditions. Considering the difficulty of obtaining labeled compound interference samples in practice, single-to-compound generalization and few-shot compound interference adaptation are investigated as two practical receiver deployment scenarios. Experimental results show that the proposed method improves the overall exact accuracy over the strongest baseline by 14.71 percentage points in single-to-compound generalization and by 14.81 percentage points in few-shot compound interference adaptation for $r=1$. △ Less

Submitted 13 July, 2026; originally announced July 2026.

Comments: Submitted to IEEE Internet of Things Journal

Articulate Intuition or Genuine Analysis? Benchmarking Epistemic Reliability in LLM-as-a-Judge Peer Reviews

Authors: Nuo Chen, Qian Wang, Qingyun Zou, Bingsheng He

Abstract: When an LLM judge calls a peer review analytical and a human committee calls another review high quality, are they tracking the same thing? We argue they are not, and that the difference matters philosophically. We operationalise Kahneman's dual-process theory into a structured rubric for peer review and release Kahneman4Review, a benchmark of 3,563 rated reviews scored along nine theoretically mo… ▽ More When an LLM judge calls a peer review analytical and a human committee calls another review high quality, are they tracking the same thing? We argue they are not, and that the difference matters philosophically. We operationalise Kahneman's dual-process theory into a structured rubric for peer review and release Kahneman4Review, a benchmark of 3,563 rated reviews scored along nine theoretically motivated textual dimensions, eight bias diagnostics, and a continuous reasoning-quality score. Three findings bear on trustworthiness: decision tier is not detectably aligned with the rubric's text-grounded epistemic-quality proxy; public-showcase agentic reviews receive higher raw scores than pooled human reviews, but length and venue explain most of the gap and the samples are not paper-paired; and ICLR review-text diagnostics shift at the 2022--2023 transition, temporally coincident with widespread LLM availability but without identifying its cause. A matched function-probe pilot further shows that the rubric distinguishes textual probes designed to contrast genuine fault-finding with surface fluency. We argue that a trustworthy reliability benchmark for LLM judges must separate analytical form from epistemic function, and propose concrete design choices toward that goal. An interactive demo is available at https://huggingface.co/spaces/nuojohnchen/Kahneman4Review. △ Less

Submitted 11 July, 2026; originally announced July 2026.

Comments: Preprint

Absence of a shell closure in $^{140}$Sn

Authors: Francesca Bonaiti, Bingcheng He, Gaute Hagen, Thomas Papenbrock

Abstract: There are conflicting theoretical results about the presence of a shell closure in the neutron-rich nucleus $^{140}$Sn. We address this controversy by performing ab initio computations, using a nuclear interaction from chiral effective field theory that accurately reproduced and predicted low-lying states in doubly magic nuclei. We verify that this interaction accurately reproduces low-lying state… ▽ More There are conflicting theoretical results about the presence of a shell closure in the neutron-rich nucleus $^{140}$Sn. We address this controversy by performing ab initio computations, using a nuclear interaction from chiral effective field theory that accurately reproduced and predicted low-lying states in doubly magic nuclei. We verify that this interaction accurately reproduces low-lying states in $^{133}$Sn. We assume that $^{140}$Sn exhibits a closed $7/2^-$ neutron subshell beyond $^{132}$Sn and compute its first excited $2^+$ state. The resulting energy is small and this contradicts the assumption. △ Less

Submitted 10 July, 2026; originally announced July 2026.

Comments: 9 pages, 3 figures

DT-Guard: Intent-Driven Reasoning-Active Training for Reasoning-Free LLM Safety Guardrail

Authors: He Liu, Changtao Miao, Xinjie Yang, Tianle Song, Yin Wu, Junchi Chen, Bintao He, Xinyuan Zhang, Bo Zhang, Shi Yan, Wei Lu, Wei Wang, Danyang Xu, Jiansheng Cai, Zhe Li

Abstract: Large language models deployed in open-world applications require safety guardrails that are both robust to complex risks and efficient enough for low-latency runtime moderation. Existing guardrails face a practical trade-off between lightweight classification-based models, which are efficient but often struggle with concealed intent, ambiguous semantics, and borderline safety decisions, and reaso… ▽ More Large language models deployed in open-world applications require safety guardrails that are both robust to complex risks and efficient enough for low-latency runtime moderation. Existing guardrails face a practical trade-off between lightweight classification-based models, which are efficient but often struggle with concealed intent, ambiguous semantics, and borderline safety decisions, and reasoning-based guards, which improve judgment quality but introduce additional token generation and inference latency. We present DT-Guard, a content safety guardrail model based on a Reasoning-Active Training, Reasoning-Free Inference paradigm. The key idea is to use reasoning supervision during training while emitting only structured safety labels at inference time. DT-Guard formulates safety judgment as a progressive decision process, Intent - Category - Safety, and constructs an intent-driven dataset with intent labels, risk categories, safety labels, and structured reasoning trajectories. To further improve hard-case robustness, we propose Rollout-Guided Progressive Hard-Case Optimization (RG-PHO), which uses multi-rollout consistency to identify stably mastered, persistently failed, and preference-unstable samples, and applies targeted supervised and preference optimization accordingly. At inference time, DT-Guard directly generates structured labels without explicit reasoning traces, preserving deployment efficiency. Experiments on prompt-side and response-side safety benchmarks show that DT-Guard achieves average F1 scores of 0.886 and 0.870, respectively. With only a 4B backbone, it reaches a dual-side average F1 of 0.878, outperforming strong 8B guardrail baselines. These results demonstrate that reasoning supervision can be effectively internalized into low-latency safety discrimination. △ Less

Submitted 7 July, 2026; originally announced July 2026.

MSCENet: A Multi-Scale Correlation Enhanced Network for Anomaly Detection

Authors: Long Zhao, Shixun Ji, Zhipeng Wang, Bin Cheng, Bin He

Abstract: In the field of multivariate time series anomaly detection, against the backdrop of increasing data complexity and complex dependencies across multiple temporal scales, traditional methods often struggle to simultaneously capture temporal dynamic features and intricate inter-series correlations. To address this, we propose an innovative framework, MSCENet, which leverages advanced spatio-temporal… ▽ More In the field of multivariate time series anomaly detection, against the backdrop of increasing data complexity and complex dependencies across multiple temporal scales, traditional methods often struggle to simultaneously capture temporal dynamic features and intricate inter-series correlations. To address this, we propose an innovative framework, MSCENet, which leverages advanced spatio-temporal learning and multi-scale learning techniques to enhance detection accuracy. MSCENet includes a fine-grained temporal convolution module that captures complex temporal dependencies through dilated convolutions, enabling the detection of both short- and long-term patterns. Additionally, the framework models inter-series relationships as a graph structure, using Mixhop graph convolutions to adaptively capture spatial dependencies across varying time scales. To support robust anomaly detection, the multi-scale gated convolution module in MSCENet integrates spatial and temporal attributes through gated mechanisms, facilitating the detection of subtle variations across multiple scales. Experimental evaluations on real-world datasets: SMD, PSM, and SWaT. It provides an adaptable and high-performance solution for anomaly detection in complex time series data environments. △ Less

Submitted 7 July, 2026; originally announced July 2026.

Weak-to-Strong Generalization via Direct On-Policy Distillation

Authors: Shiyuan Feng, Huan-ang Gao, Haohan Chi, Hanlin Wu, Zhilong Zhang, Zheng Jiang, Bingxiang He, Wei-Ying Ma, Ya-Qin Zhang, Hao Zhou

Abstract: Reinforcement learning with verifiable rewards (RLVR) is a powerful recipe for improving language-model reasoning, but it is expensive to repeat on every new strong model because the target model must generate many rollouts during training. As models scale, post-training itself becomes a bottleneck. We study a weak-to-strong alternative: run RL on a smaller model where rollouts are cheaper, then r… ▽ More Reinforcement learning with verifiable rewards (RLVR) is a powerful recipe for improving language-model reasoning, but it is expensive to repeat on every new strong model because the target model must generate many rollouts during training. As models scale, post-training itself becomes a bottleneck. We study a weak-to-strong alternative: run RL on a smaller model where rollouts are cheaper, then reuse what that RL run learned to improve a stronger target model. Directly distilling the post-RL weak teacher is not enough, because the teacher's final policy mixes useful RL gains with the limitations of the smaller model. We propose Direct On-Policy Distillation (Direct-OPD), which transfers the teacher's RL-induced policy shift instead. Direct-OPD compares the post-RL teacher with its own pre-RL reference and treats their log-ratio as a dense implicit reward for the student. In plain terms, the checkpoint pair tells us which actions RL made the weak model more or less likely to take, and Direct-OPD applies that signal on the stronger student's own on-policy states. This directly reuses the weak model's RL supervision signal without running sparse-reward RL on the target model. Empirically, Direct-OPD consistently leverages weaker teachers to improve stronger target models; notably, it boosts Qwen3-1.7B from 48.3% to 58.3% on AIME 2024 in just 4 hours on 8 A100 GPUs. It outperforms step-matched direct RL and enables the sequential composition of multiple policy shifts. Our results show that RL outcomes can be reused across model scales as implicit reward signals, not merely as final models to imitate. △ Less

Submitted 8 July, 2026; v1 submitted 6 July, 2026; originally announced July 2026.

Comments: Project Page: https://bytedtsinghua-sia.github.io/Direct-OPD/

Branch-JEPA: Finite-Support Predictive Distributions for JEPA World Models

Authors: Zhi Song, Ximing Xing, Zhenchao Tang, hanbo Huang, Jiehui Huang, Weilong Yan, Tianxu Lv, Minghao Yang, Zhongzheng Niu, Bing He, Lusheng Wang, Jianhua Yao

Abstract: Joint-embedding predictive architectures (JEPAs) learn dynamics by predicting future observations in representation space. Yet most JEPA world models return one latent successor, even when hidden intent, partial observation, or stochastic dynamics make several futures plausible. We introduce Branch-JEPA, which replaces this point-valued transition with a context-weighted finite set of latent succe… ▽ More Joint-embedding predictive architectures (JEPAs) learn dynamics by predicting future observations in representation space. Yet most JEPA world models return one latent successor, even when hidden intent, partial observation, or stochastic dynamics make several futures plausible. We introduce Branch-JEPA, which replaces this point-valued transition with a context-weighted finite set of latent successors. Every branch is decoded independently, and the complete set is retained at inference. The architecture supports two complementary training regimes: specialization for recovering separated successors and full-set Energy-Score training for distributional fidelity. In a locked five-seed evaluation on the Argoverse~2 official validation split, full-set training improves trajectory Energy Score by $5.8$--$6.5\%$ and probability-weighted trajectory distance by $9.3$--$10.4\%$ over matched-$K{=}6$ assignment and transport objectives, while retaining $5.36$ endpoint-deduplicated effective branches. In a parameter-exact official-validation comparison, latent branching retains $10.3\%$ more effective modes and improves Energy Score, expected ADE, and Brier in all five paired seeds over branching only at the output decoder; every paired 95\% interval excludes zero. In an OGBench graph audit, Branch-JEPA increases teleport verified-route existence to $19.2\%$ versus $3.9\%$ for the MDN. Its raw-support advantage also persists with 29-D state and RGB observations. Together, latent branching preserves more distinct futures, while full-set scoring improves the quality of the resulting predictive distribution. △ Less

Submitted 3 August, 2026; v1 submitted 6 July, 2026; originally announced July 2026.

Optimal Mixture-of-Experts Model Averaging for Conditional Generative Models

Authors: Shijin Gong, Baihua He, Xinyu Zhang

Abstract: Conditional generative models have emerged as powerful tools for sampling from target conditional distributions, driving substantial advances across a wide range of scientific and applied domains. As these models proliferate, practitioners often face multiple plausible generators whose performance can vary with the task, data, or input condition. We propose an optimal model averaging framework for… ▽ More Conditional generative models have emerged as powerful tools for sampling from target conditional distributions, driving substantial advances across a wide range of scientific and applied domains. As these models proliferate, practitioners often face multiple plausible generators whose performance can vary with the task, data, or input condition. We propose an optimal model averaging framework for conditional generative models, allowing candidate generators to be combined even when they are accessible only through conditional samples without tractable densities. Specifically, we use a sample-based maximum mean discrepancy between conditional distributions, which first leads to a static model averaging method, StaticMA, assigning fixed weights to different candidates. In addition, we develop MoEMA (mixture-of-experts model averaging), an input-adaptive method that parameterizes covariate-dependent weights through a softmax neural-network gate. We establish in-sample and out-of-sample asymptotic optimality for the proposed methods, together with consistency of the estimated adaptive weight function under regularity conditions. The framework applies directly to Euclidean responses and extends to unstructured data by combining our formulation with fixed representation maps. Across a broad set of simulations and real-data studies spanning tabular, image, and text modalities, MoEMA generally improves over competing baselines, demonstrating the effectiveness of our proposed methods. △ Less

Submitted 5 July, 2026; originally announced July 2026.

Comments: 32 pages, 1 figures

Conflict-Based Lazy Search for Fast Multi-Manipulator Planning

Authors: Dongliang Zheng, Zhipeng Wang, Siqi Wang, Yuxi Lu, Bin He, Hesheng Wang, Panagiotis Tsiotras

Abstract: Employing multiple manipulators can boost efficiency and accomplish tasks that a single manipulator cannot do. However, real-time planning for multiple manipulators in a cluttered workspace still poses significant challenges for planning algorithms. This article proposes a new planning algorithm called Conflict-Based Lazy Search (CBLS) for multimanipulator planning. CBLS is built on Conflict-Based… ▽ More Employing multiple manipulators can boost efficiency and accomplish tasks that a single manipulator cannot do. However, real-time planning for multiple manipulators in a cluttered workspace still poses significant challenges for planning algorithms. This article proposes a new planning algorithm called Conflict-Based Lazy Search (CBLS) for multimanipulator planning. CBLS is built on Conflict-Based Search (CBS), an efficient multiagent pathfinding (MAPF) algorithm that has shown an order of magnitude speedup over previous approaches [1], [2]. CBS addresses MAPF by solving many single-agent pathfinding (SAPF) problems. Thus, its planning time directly depends on the efficiency of the SAPF algorithm adopted. Our CBLS algorithm enhances CBS with precomputation and lazy search. First, a lazily evaluated graph with controlled sparsity is precomputed for a single manipulator. Second, we propose the Lazy Edged-based A* (LEA*) for efficient SAPF. Since edge evaluation is the computational bottleneck of manipulator planning, LEA* uses lazy search and an edge queue to reduce the number of edge evaluations. We show that LEA* is optimally vertex efficient and has improved edge efficiency compared to A*. We apply the proposed CBLS to multi-manipulator planning problems and show its superior performance by comparing it with CBS and a sampling-based algorithm, namely, RRT-Connect. △ Less

Submitted 5 July, 2026; originally announced July 2026.

Journal ref: IEEE/ASME Transactions on Mechatronics(TMECH) 2026

Enhanced Feature Extraction for IoT Network Intrusion Detection Using GNNs and KAN

Authors: Long Zhao, Shixun Ji, Bin Cheng, Bin He

Abstract: Recent advancements in the Internet of Things (IoT) emphasize the urgent need for advanced network security, as IoT networks feature dynamic topologies, imbalanced traffic, and complex attack patterns. Unlike general IT networks, IoT environments exhibit extreme heterogeneity and sparse topologies. Traditional GNN-based intrusion detection methods often struggle to efficiently model node and edge… ▽ More Recent advancements in the Internet of Things (IoT) emphasize the urgent need for advanced network security, as IoT networks feature dynamic topologies, imbalanced traffic, and complex attack patterns. Unlike general IT networks, IoT environments exhibit extreme heterogeneity and sparse topologies. Traditional GNN-based intrusion detection methods often struggle to efficiently model node and edge features or capture fine-grained anomalies in such settings. To address this, we propose SKGFusionKAN, a novel IoT-tailored approach enhancing GraphSAGE with a multi-scale selective kernel attention mechanism. This enables adaptive extraction of node and edge features under diverse traffic conditions. Specifically, our edge-oriented message passing strengthens information propagation, while selective kernel attention adaptively weights edge-derived information from different scales to handle heterogeneity. We also introduce a gated fusion process to dynamically integrate multi-scale features, improving robustness against evolving attacks. Finally, we leverage Kolmogorov-Arnold Networks (KAN) for classification, offering superior nonlinear modeling capabilities essential for detecting intricate, low-frequency attacks. To our knowledge, this work presents a comprehensive integration of GNNs and KAN with dedicated architectural innovations for IoT intrusion detection. Extensive experiments on four NIDS benchmarks show that SKGFusionKAN consistently outperforms state-of-the-art approaches in binary and multiclass tasks, demonstrating its potential for IoT security. △ Less

Submitted 3 July, 2026; originally announced July 2026.

$l$-forbidden $\mathbf{M1}$ strengths near $^{100}$Sn from knockout reactions in Cd and Sn

Authors: T. J. Gray, K. L. Jones, R. Grzywacz, B. A. Brown, A. Gade, B. C. He, T. Miyagi, A. Peter, M. J. Basson, T. Beck, C. M. Campbell, G. Cerizza, J. Chung-Jung, I. Cox, P. Farris, R. Ghimire, S. Gillespie, M. Grinder, A. Hill, S. D. Pain, A. Palmisano-Kyle, K. P. Rykaczewski, D. Weisshaar, M. Williams

Abstract: Neutron knockout reactions on beams of $^{104,102}$Cd, and $^{104}$Sn are presented. States in the residual $^{103,101}$Cd and $^{103}$Sn nuclei are populated, including low-lying $7/2^+$ states of $νg_{7/2}$ character. These states have half-lives $\approx 400$ ps due to their low energy and hindered $B(M1; 7/2^+ \rightarrow 5/2^+)$ strengths. The excited-state half-lives were measured using thei… ▽ More Neutron knockout reactions on beams of $^{104,102}$Cd, and $^{104}$Sn are presented. States in the residual $^{103,101}$Cd and $^{103}$Sn nuclei are populated, including low-lying $7/2^+$ states of $νg_{7/2}$ character. These states have half-lives $\approx 400$ ps due to their low energy and hindered $B(M1; 7/2^+ \rightarrow 5/2^+)$ strengths. The excited-state half-lives were measured using their Doppler-shifted lineshapes, and the resulting $B(M1)$ strengths are compared to Valence Space In Medium Similarity Renormalization Group (VS-IMSRG) calculations. The VS-IMSRG calculations under-predict the $l$-forbidden $M1$ strengths in the $^{100}$Sn region, as well as in other regions of the nuclear chart near $^{40}$Ca and $^{208}$Pb. △ Less

Submitted 2 July, 2026; originally announced July 2026.

Comments: 9 pages + 1 page supplementary material, 13 figures, accepted in PRC

Beyond Document Grounding: Span-Level Hallucination Detection over Code, Tool Output, and Documents

Authors: Ádám Kovács, Bowei He, Xue Liu, István Boros, Szilveszter Tóth, Gábor Recski

Abstract: Hallucination detection for retrieval-augmented generation (RAG) is usually evaluated on natural-language document evidence. However, grounded generation systems increasingly rely on structured inputs: source code, developer-tool output, markdown documents, tables, and repository metadata. We introduce a unified benchmark for span-level hallucination detection over code, tool output, structured do… ▽ More Hallucination detection for retrieval-augmented generation (RAG) is usually evaluated on natural-language document evidence. However, grounded generation systems increasingly rely on structured inputs: source code, developer-tool output, markdown documents, tables, and repository metadata. We introduce a unified benchmark for span-level hallucination detection over code, tool output, structured documents, and existing natural-language RAG datasets. The benchmark is built by starting from grounded correct answers, injecting localized hallucinations with exact character labels, and validating the code test split with evidence-based review. Our fine-tuned Qwen3.5-2B detector reaches 0.689 span-F1 on the unified test set and 0.60 on the code-agent source, where it substantially outperforms LettuceDetect-large (0.17) and the strongest zero-shot LLM judges we evaluated (at most 0.22). The same model remains competitive on established natural-language benchmarks, with 81.8 RAGTruth example-F1 and 0.724 English PsiloQA IoU. △ Less

Submitted 1 July, 2026; originally announced July 2026.

Comments: 8 pages

Multi-Label Node Classification with Label Influence Propagation

Authors: Yifei Sun, Zemin Liu, Bryan Hooi, Yang Yang, Rizal Fathony, Jia Chen, Bingsheng He

Abstract: Graphs are a complex and versatile data structure used across various domains, with possibly multi-label nodes playing a particularly crucial role. Examples include proteins in PPI networks with multiple functions and users in social or e-commerce networks exhibiting diverse interests. Tackling multi-label node classification (MLNC) on graphs has led to the development of various approaches. Some… ▽ More Graphs are a complex and versatile data structure used across various domains, with possibly multi-label nodes playing a particularly crucial role. Examples include proteins in PPI networks with multiple functions and users in social or e-commerce networks exhibiting diverse interests. Tackling multi-label node classification (MLNC) on graphs has led to the development of various approaches. Some methods leverage graph neural networks (GNNs) to exploit label co-occurrence correlations, while others incorporate label embeddings to capture label proximity. However, these approaches fail to account for the intricate influences between labels in non-Euclidean graph data. To address this issue, we decompose the message passing process in GNNs into two operations: propagation and transformation. We then conduct a comprehensive analysis and quantification of the influence correlations between labels in each operation. Building on these insights, we propose a novel model, Label Influence Propagation (LIP). Specifically, we construct a label influence graph based on the integrated label correlations. Then, we propagate high-order influences through this graph, dynamically adjusting the learning process by amplifying labels with positive contributions and mitigating those with negative influence. Finally, our framework is evaluated on comprehensive benchmark datasets, consistently outperforming SOTA methods across various settings, demonstrating its effectiveness on MLNC tasks. △ Less

Submitted 1 July, 2026; originally announced July 2026.

Comments: Accepted to ICLR 2025

Measurement-Based Characterization and Statistical Modeling of 6G Urban Low-Altitude A2G Channels across FR1 and FR3

Authors: Bin Ao, Jianhua Zhang, Pan Tang, Peijie Liu, Boyang He, Hao Zheng

Abstract: Unmanned aerial vehicle (UAV) communications have been recognized as a key component of future sixth-generation (6G) space-air-ground-sea integrated networks. Accurate characterization and modeling of air-to-ground (A2G) channels are essential for the design and optimization of low-altitude communication systems. This paper presents a wideband A2G channel measurement campaign in an urban environme… ▽ More Unmanned aerial vehicle (UAV) communications have been recognized as a key component of future sixth-generation (6G) space-air-ground-sea integrated networks. Accurate characterization and modeling of air-to-ground (A2G) channels are essential for the design and optimization of low-altitude communication systems. This paper presents a wideband A2G channel measurement campaign in an urban environment at 2.85 and 4.6~GHz in FR1 and 7.25~GHz in the FR3 frequency band, each with a bandwidth of 250~MHz. To enable reliable line-of-sight (LoS) and non-line-of-sight (NLoS) propagation state identification, a weakly supervised method is developed by fusing geometric priors, channel features, and spatial consistency constraints. Furthermore, based on the measured data, A2G channel characteristics are extracted and analyzed under LoS/NLoS conditions across different frequency bands, including path loss (PL), shadow fading (SF), power delay profile, root-mean-square delay spread (RMS-DS), and Rician $K$-factor. The results show that the close-in model fits the measured PL more accurately than the 3GPP reference model, and that NLoS propagation leads to larger path loss exponents and stronger SF than LoS propagation. For channel delay characteristics, higher-frequency channels exhibit fewer effective MPCs and weaker delay dispersion, indicating increased channel sparsity. Specifically, the mean RMS-DS under LoS conditions decreases from 93.11 to 46.84~ns, while the mean Rician $K$-factor increases from 9.16 to 12.88~dB. The statistical results further show that the RMS-DS and the Rician $K$-factor can be well characterized by lognormal and normal distributions, respectively. Moreover, the movement of the receiver in a complex scattering environment intensifies the spatial non-stationarity of the A2G channel. △ Less

Submitted 1 July, 2026; originally announced July 2026.

Comments: 12 pages, 10 figures, 36 conferences

Scalable Behaviour Cloning on Browser Using via Skill Distillation

Authors: Kaisen Yang, Zheng Jiang, Yuzhao Peng, Houde Qian, Boshi Zhang, Youjie Zheng, Shijin Hong, Qingle Liu, Ruoyu Han, Bohan Lyu, Bingxiang He, Eren Cai, Calvin Xiao, Qinhuai Na

Abstract: Internet users collectively perform an enormous range of skilled work through web browsers, from software development and document editing to search, forms, and enterprise workflows, making human browsing a highly scalable but under-exploited source of reusable browser skills. We argue that the bottleneck for browser agents is decision-making under incomplete information rather than low-level oper… ▽ More Internet users collectively perform an enormous range of skilled work through web browsers, from software development and document editing to search, forms, and enterprise workflows, making human browsing a highly scalable but under-exploited source of reusable browser skills. We argue that the bottleneck for browser agents is decision-making under incomplete information rather than low-level operation, and that the priors agents lack are already implicit in human interaction traces. We therefore study scalable behavior cloning for browser agents via skill distillation, converting user interaction trajectories into compact natural-language skills that agents can read, retrieve, reuse, and compose directly. We further organize the distilled skills into a skill graph so that growth proceeds through consolidation rather than unbounded accumulation. This suggests that the scalability of browser agents may come less from manually designed tasks and more from the collective skills already expressed by internet users. Our project is available at: https://lab.einsia.ai/browserbc/. △ Less

Submitted 30 June, 2026; originally announced June 2026.

A Survey of Physical-layer Authentication Enhanced by Emerging Spatial Domain Technologies

Authors: Yuhao Chen, Boxiang He, Junshan Luo, Shilian Wang, Yiyan Ma, Hao Xu, Lei Yao, Jing Lei, Arumugam Nallanathan, Kai-Kit Wong

Abstract: This article surveys spatial-domain-enhanced Physical-layer Authentication (PLA), with Dual-polarized Antennas (DPA), Massive Multiple-Input Multiple-Output (MIMO), and Reconfigurable Intelligent Surfaces (RIS) as the primary focus. With the rapid growth of wireless deployments, authentication mechanisms face stringent requirements for high security, low overhead, and low latency. PLA offers light… ▽ More This article surveys spatial-domain-enhanced Physical-layer Authentication (PLA), with Dual-polarized Antennas (DPA), Massive Multiple-Input Multiple-Output (MIMO), and Reconfigurable Intelligent Surfaces (RIS) as the primary focus. With the rapid growth of wireless deployments, authentication mechanisms face stringent requirements for high security, low overhead, and low latency. PLA offers lightweight identity verification by exploiting physical-layer characteristics. However, the effectiveness of PLA critically depends on how physical observations are constructed and validated under wireless channels. Unlike existing surveys that mainly organize PLA by authentication modality, feature source, and evaluation metrics, this work emphasizes the connection between spatial-domain enhancement mechanisms, the resulting feature representation, and the authentication procedure. We review how DPA, Massive MIMO, and RIS reshape PLA feature representation, and we summarize newly introduced security threats along with representative defense strategies. Case studies further illustrate the practical impact, such as representative detection-probability trends across Signal-to-Noise Ratio regimes and quantitative comparisons among representative schemes. Finally, we outline promising future opportunities enabled by Dynamic Metasurface Antennas, Extra-large MIMO, and spatial configuration with artificial intelligence. △ Less

Submitted 27 June, 2026; originally announced June 2026.

Cross-channel Specific Emitter Identification and Verification via Signal Envelope

Authors: Yuhao Chen, Boxiang He, Shilian Wang, Jing Lei

Abstract: Specific emitter identification (SEI) determines which known emitter a received signal originates from, while specific emitter verification (SEV) determines whether the received signal genuinely comes from its claimed emitter. In this paper, we consider the effect of wireless fading channels on SEI and SEV. When the Rician $K$-factor varies, the resulting distribution shift induced by the channel… ▽ More Specific emitter identification (SEI) determines which known emitter a received signal originates from, while specific emitter verification (SEV) determines whether the received signal genuinely comes from its claimed emitter. In this paper, we consider the effect of wireless fading channels on SEI and SEV. When the Rician $K$-factor varies, the resulting distribution shift induced by the channel degrades both identification and verification performance. To address this issue, we first theoretically prove that the coefficient of variation of the signal envelope is strictly monotonic with respect to the Rician $K$-factor. Motivated by this property, we propose an envelope-guided adaptive feature modulation (EAFM) identifier for SEI and an EAFM with Mahalanobis distance metric learning (EAFM-MD) verifier for SEV. Specifically, the proposed EAFM identifier adopts a dual-branch neural network to extract device-oriented features from the IQ-domain input and channel-conditioning features from the normalized signal envelope, and adaptively modulates the former via feature-wise linear modulation. Then, we extend the EAFM identifier to an EAFM-MD verifier. The device-fingerprint library is constructed by storing the feature centroid and covariance for each enrolled device, along with the within-device Mahalanobis distances of training signals. For verification, the Mahalanobis distance between the extracted test features and each stored centroid is computed using the stored covariance matrix, and the minimum distance is compared to the corresponding device threshold to make a decision. Finally, numerical results show that the proposed EAFM identifier improves cross-channel identification performance, while the proposed EAFM-MD verifier achieves superior detection performance against unknown spoofing attacks. △ Less

Submitted 27 June, 2026; originally announced June 2026.

UniFormer: Efficient and Unified Model-Centric Scaling for Industrial Recommendation

Authors: Bo Chen, Jinlong Jiao, Tijian Hu, Ruihao Zhang, Yanzhi Liu, Chenghou Jin, Qinglin Jia, Baixuan He, Hechang Pan, Yiwu Liu, Jian Liang, Chaoyi Ma, Ruiming Tang, Han Li, Kun Gai

Abstract: Recently, substantial progress has been made in industrial recommendation through component-centric model scaling, where individual components such as behavior modeling, feature interaction, or task modeling are independently scaled to improve model capacity. Although recent methods such as HyFormer and OneTrans further explore cross-module co-scaling by jointly modeling behavior and interaction,… ▽ More Recently, substantial progress has been made in industrial recommendation through component-centric model scaling, where individual components such as behavior modeling, feature interaction, or task modeling are independently scaled to improve model capacity. Although recent methods such as HyFormer and OneTrans further explore cross-module co-scaling by jointly modeling behavior and interaction, their designs are still confined to the feature space and lack a unified model-centric scaling framework over the overall modeling space. In this paper, we propose UniFormer, an efficient and unified model-centric scaling framework for industrial recommender systems. To improve efficiency, UniFormer decomposes the overall modeling space into feature and task spaces, which are modeled by stacked Feature-space Interaction Modules and Task-space Interaction Modules, respectively. Moreover, UniFormer introduces semantic-based tokenization scheme to enable user-item decoupling, thereby achieving request-level inference acceleration. To prevent preference collapse, UniFormer employs multi-sequence cross-attention to separately capture heterogeneous behavior patterns, followed by the self-attention to enhance interaction modeling. Besides, dedicated multi-view FFNs are introduced to support flexible and scalable parameter scaling across different modeling components. Extensive online A/B testing in two production scenarios, Kuaishou and Kuaishou Lite, shows that UniFormer consistently improves user engagement and interaction metrics, achieving gains of +0.101%/+0.260% in App Stay Time and +0.729%/+1.113% in Watch Time, respectively. △ Less

Submitted 25 June, 2026; originally announced June 2026.

IDEA: Insensitive to Dynamics Mismatch via Effect Alignment for Sim-to-Real Transfer in Multi-Agent Control

Authors: Chenlong Liu, Zhuohui Zhang, Xinyan Chen, Zhipeng Wang, Bin Cheng, Bin He

Abstract: Complex multi-agent control tasks remain challenging for traditional rule-based and model-based approaches, motivating the adoption of learning-based methods. However, learning-based methods often struggle with sim-to-real transfer because they rely on accurate dynamics modeling or system identification and learn policies in low-level control spaces that are highly sensitive to dynamics mismatch,… ▽ More Complex multi-agent control tasks remain challenging for traditional rule-based and model-based approaches, motivating the adoption of learning-based methods. However, learning-based methods often struggle with sim-to-real transfer because they rely on accurate dynamics modeling or system identification and learn policies in low-level control spaces that are highly sensitive to dynamics mismatch, making them costly and fragile in complex environments. To address this issue, we propose a sim-to-real method for multi-agent control, which is insensitive to dynamics mismatch via effect alignment. Our method combines random environmental structure with discrete semantic actions through closed-loop control, elevating policy learning to a semantic abstraction level. Additionally, we develop an action synchronization mechanism that mitigates inter-agent action timing mismatches, thereby enhancing the temporal consistency of the system. Experiments on four multi-agent navigation tasks demonstrate that our method substantially improves training efficiency over mainstream transfer methods and achieves higher success rates in real-world scenarios, thereby improving the robustness and deployment stability of multi-agent systems under dynamics mismatch. △ Less

Submitted 24 June, 2026; originally announced June 2026.

Comments: 8 pages, 6 figures

ForceBand: Learning Forceful Manipulation with sEMG

Authors: Botao He, Zhi Wang, Linna Kuang, Ishaan Ghosh, Jitendra Malik, Cornelia Fermuller, Tingfan Wu, Jiayuan Mao, Ruoshi Liu, Haozhi Qi, Yiannis Aloimonos

Abstract: Human demonstrations are a scalable data source for learning robot manipulation policies. However, common sources of human demonstration data, such as motion-capture trajectories and internet videos, capture mostly motion and appearance while missing the contact forces that are critical for force-sensitive manipulation. In this paper, we introduce ForceBand, a low-cost wrist-worn sEMG system that… ▽ More Human demonstrations are a scalable data source for learning robot manipulation policies. However, common sources of human demonstration data, such as motion-capture trajectories and internet videos, capture mostly motion and appearance while missing the contact forces that are critical for force-sensitive manipulation. In this paper, we introduce ForceBand, a low-cost wrist-worn sEMG system that turns human muscle activity into force-enriched demonstrations. We first collect a 10-hour multimodal dataset containing egocentric video, sEMG, IMU, and fingertip force measurements across diverse actions and objects. Using this dataset, we pre-train an EMG2Force model that predicts per-finger forces from sEMG and IMU signals. After a short user-specific calibration, users can collect target-task demonstrations using only ForceBand and video; EMG2Force then labels these demonstrations with per-finger force traces, producing force-augmented demonstrations for robot policy learning. Experiments show that ForceBand recovers fine-grained fingertip interactions with over 50% lower force prediction error than vision-based baselines and achieves an 87% success rate on pick, squeeze, and place tasks that require object-specific force control across objects with diverse shapes, sizes, and weights. Project website: https://forceband-emg.github.io △ Less

Submitted 24 June, 2026; originally announced June 2026.

Broadband multiwavelength properties of the archetypal blazar 3C 279 during the 2017 Event Horizon Telescope campaign

Authors: G. Principe, J. C. Algaba, E. Aviano, W. Y. Cheong, K. Hada, D. Haggard, A. Hahn, S. G. Jorstad, E. V. Kravchenko, Y. Kovalev, S. S. Lee, M. Lisakov, S. Markoff, A. P. Marscher, M. Sasada, P. Voitsik, Kazunori Akiyama, Ezequiel Albentosa-Ruiz, Antxon Alberdi, Walter Alef, Richard Anantua, Eleni Antonopoulou, Keiichi Asada, Rebecca Azulay, Anne-Kathrin Baczko , et al. (508 additional authors not shown)

Abstract: The archetypal blazar 3C 279 hosts a prominent relativistic jet and exhibits strong broadband variability across the electromagnetic spectrum. In April 2017, the Event Horizon Telescope (EHT) observed 3C 279, alongside one of the most extensive quasi-simultaneous multiwavelength (MWL) campaigns ever conducted. With the aim of investigating the physical processes governing 3C 279, we analyzed indiv… ▽ More The archetypal blazar 3C 279 hosts a prominent relativistic jet and exhibits strong broadband variability across the electromagnetic spectrum. In April 2017, the Event Horizon Telescope (EHT) observed 3C 279, alongside one of the most extensive quasi-simultaneous multiwavelength (MWL) campaigns ever conducted. With the aim of investigating the physical processes governing 3C 279, we analyzed individual observations and multiband light curves, and constructed a new quasi-simultaneous MWL spectrum. We also performed phenomenological modeling using the turbulent extreme multi-zone (TEMZ) model to constrain the fundamental physical properties of the source. The EHT observations reveal a clear flux increase in the innermost core between April 5 and 11, 2017. Over a broader timescale, radio measurements at longer wavelengths show concurrent enhancements in core flux and polarization around mid-April, coinciding with the ejection of a superluminal knot. Record UV-optical flares with strong polarization variability occurred in late March, followed by gamma-ray activity that declined before the end of the EHT observing period. During this interval, the source remained in a low X-ray state and showed no detectable VHE emission. The TEMZ modeling suggests that the broadband spectrum and variability of 3C 279 can be explained within a jet scenario in which turbulent plasma cells are compressed by a stationary conical shock. However, alternative interpretations, such as magnetic reconnection or a moving shock-in-jet event, remain plausible. This coordinated MWL campaign advances our understanding of the origin of jet and gamma-ray emission in 3C 279, while also providing a comprehensive publicly available dataset that will serve as a valuable reference for future studies. △ Less

Submitted 24 June, 2026; originally announced June 2026.

Comments: 29 pages, 13 figures, accepted by Astronomy & Astrophysics on June 18, 2026

Internal Data Repetition Destroys Language Models

Authors: Jessica Chudnovsky, Joshua Kazdan, Noam Levi, Rylan Schaeffer, Yegor Denisov-Blanch, Bo He, Mehmet Donmez, Sanmi Koyejo, David Donoho

Abstract: Language models are running out of high-quality training data, and even aggressively deduplicated corpora retain some amount of repetition. Earlier controlled studies predated Chinchilla-style scaling laws and could only measure the cost of repetition indirectly. We revisit repetition in the Chinchilla era, using a fitted no-repetition scaling law to report Compute-Equivalent Gain and Compute-Equi… ▽ More Language models are running out of high-quality training data, and even aggressively deduplicated corpora retain some amount of repetition. Earlier controlled studies predated Chinchilla-style scaling laws and could only measure the cost of repetition indirectly. We revisit repetition in the Chinchilla era, using a fitted no-repetition scaling law to report Compute-Equivalent Gain and Compute-Equivalent Loss. We show that under this modernized paradigm, repetition damage is systematic in three ways. First, holding compute allocated to repeated data constant, eval loss peaks at an intermediate repeat count $\Rep$; repeating a moderately sized subset a moderate number of times damages performance more than repeating a large subset a few times or a small subset many times. Second, the location of this peak is well-fit by a power law in model size; this scaling law reveals that the most damaging number of repeated data grows more quickly than compute. Finally, when repeated documents consume 10\% of the FLOPs budget in a controlled exact-document repetition setting, the compute-equivalent loss can be large: on FineWeb-Edu-Dedup, the most damaging repeat count for a Qwen3-style 344M-parameter model at $\OT=1$ matches the loss of a no-repetition run using 67% of the FLOPs. We demonstrate that these phenomena are not language-model-specific, and can be analytically understood in a simple statistical model: a misspecified linear regression with verbatim duplicates reproduces the same qualitative loss peak, quantifying how such peaks can arise from a statistical tradeoff between memorization and generalization. Our findings add precision to the study of duplication in language models, allowing practitioners to quantify the wasted compute incurred by the presence and repeat structure of duplicates in pretraining corpora. △ Less

Submitted 23 June, 2026; originally announced June 2026.

NatureBench: Can Coding Agents Match the Published SOTA of Nature-Family Papers?

Authors: Yuru Wang, Lejun Cheng, Yuxin Zuo, Sihang Zeng, Bingxiang He, Che Jiang, Junlin Yang, Yuchong Wang, Kaikai Zhao, Weifeng Huang, Kai Tian, Zhenzhao Yuan, Jincheng Zhong, Weizhi Wang, Ning Ding, Bowen Zhou, Kaiyan Zhang

Abstract: We introduce NatureBench, a cross-discipline benchmark of 90 tasks distilled from peer-reviewed Nature-family publications, designed to evaluate whether AI coding agents can move beyond reproduction toward discovery on real scientific problems. NatureBench is built on NatureGym, an automated pipeline that constructs a standardized, per-task containerized environment from a source paper, addressing… ▽ More We introduce NatureBench, a cross-discipline benchmark of 90 tasks distilled from peer-reviewed Nature-family publications, designed to evaluate whether AI coding agents can move beyond reproduction toward discovery on real scientific problems. NatureBench is built on NatureGym, an automated pipeline that constructs a standardized, per-task containerized environment from a source paper, addressing the environment-fragmentation problem that has limited the credibility of prior agent-on-research benchmarks. Evaluating ten frontier agent configurations under a strict web-search-disabled protocol, we find that the strongest model surpasses SOTA on only 17.8% of tasks under the g>0.1 criterion. Analysis of method pathways reveals that agents succeed primarily through methodological translation, converting scientific tasks into familiar supervised prediction problems, rather than through genuine scientific invention. Failures are dominated by wrong method choice and insufficient compute budget, not by task misunderstanding. We release the benchmark, the NatureGym pipeline, and a public leaderboard with maintainer-side reproduction. Code: https://github.com/FrontisAI/NatureBench △ Less

Submitted 6 July, 2026; v1 submitted 23 June, 2026; originally announced June 2026.

Comments: Add results of GLM-5.2 and MinMax-M3

Observation of stopping power reduction at strong ion-plasma coupling

Authors: Yun Liu, Jieru Ren, Zhigang Deng, Wei Qi, Bubo Ma, Wenqing Wei, Shizheng Zhang, Xuyang Luo, Ziqian Zhao, Mingzhe Yang, Yifang Gao, Xueguang Ren, Jianxing Li, Dieter H. H. Hoffmann, Xing Wang, Zhongfeng Xu, Shaoyi Wang, Quanping Fan, Bo Cui, Weiwu Wang, Sixin Wu, Yue Yang, Zhurong Cao, Zongqing Zhao, Yuqiu Gu , et al. (8 additional authors not shown)

Abstract: Ion stopping in dense plasma is crucial for stellar evolution and fusion ignition. However, its behavior in the strong ion-plasma coupling regime beyond the linear limit has long remained elusive, due to formidable experimental challenges. Here we report the first experimental investigation of ion stopping at an unprecedented coupling parameter exceeding unity, achieved by sending laser-accelerate… ▽ More Ion stopping in dense plasma is crucial for stellar evolution and fusion ignition. However, its behavior in the strong ion-plasma coupling regime beyond the linear limit has long remained elusive, due to formidable experimental challenges. Here we report the first experimental investigation of ion stopping at an unprecedented coupling parameter exceeding unity, achieved by sending laser-accelerated short-pulse and intense quasi-monoenergetic carbon ions ($\sim$583 keV/u, C$^{5+}$) into a uniform, long-lived, well-characterized dense plasma target ($T_e$ $\approx$ 17 eV, $n_e$ $\approx$ 4$\times$10$^{20}$ cm$^{-3}$). By simultaneously measuring ion energy loss and charge-state evolution, we eliminated key experimental ambiguities arising from charge-state determination. Our results clearly show a reduction in stopping power compared with predictions from standard linear dielectric response or binary collision models, and they agree well with the hybrid calculation of molecular dynamics with quantum corrections. The importance of nonlinear screening effects arising from many-body interactions and quantum effects due to the wave nature of electrons was demonstrated at strong coupling. This work establishes a definitive high-fidelity experimental benchmark for collisional dynamics in the strong-coupling regime. It offers critical insight for accurate modeling of energy transport in inertial confinement fusion and astrophysical plasmas. △ Less

Submitted 15 July, 2026; v1 submitted 22 June, 2026; originally announced June 2026.

UniFS: Unified Fast-to-Slow Hierarchical Architecture for Vision-Language-Action Models

Authors: Lin Sun, Zhiwei Guan, Conglin Wang, Zihong Chen, Jianhai Yu, Zongsheng Li, Boyong He, Tao Sun, Jiale Cao, Lige Liu

Abstract: Mainstream Fast-Slow dual system vision-language-action models decouple a high-frequency action expert from a low-frequency vision-language model for efficiency, yet they face a fundamental frequency dilemma: large update gaps cause semantic drift from stale context, while small gaps erode the intended computational savings. Moreover, because the action expert receives only the VLM's final-layer r… ▽ More Mainstream Fast-Slow dual system vision-language-action models decouple a high-frequency action expert from a low-frequency vision-language model for efficiency, yet they face a fundamental frequency dilemma: large update gaps cause semantic drift from stale context, while small gaps erode the intended computational savings. Moreover, because the action expert receives only the VLM's final-layer representation at a single fixed frequency, rich intermediate features are discarded, limiting both information coupling and manipulation precision. Inspired by multi-timescale neural processing in the human brain, we introduce UniFS, a unified fast-to-slow architecture that resolves these challenges through three key designs. First, we stratify the VLM layers into groups with progressively decreasing update frequencies, enabling shallow layers to capture fast-changing dynamics while deeper layers cache stable semantic context. Second, a latent vector inversion mechanism re-routes the interaction order between multi-scale VLM features and the action expert, aligning fast-varying representations with fine-grained action decoding and slow-varying ones with coarse planning. Third, a multi-level supervision strategy enforces a coarse-to-fine learning hierarchy across temporal scales. Together, these designs enable richer cross-frequency information transfer within a single backbone, while the low-frequency pathways additionally preserve temporal context across steps. Experiments on LIBERO show that UniFS achieves state-of-the-art performance (98.3\% average success rate, a 2.5\% gain over VLA-Adapter baseline) while reducing average inference latency from 36.5~ms to 17.8~ms (2.1$\times$ speedup). Real-robot experiments on a Franka platform further validate its practical applicability. Code is opensourced at https://github.com/linsun449/UniFS. △ Less

Submitted 21 June, 2026; originally announced June 2026.

Comments: Code is opensourced at https://github.com/linsun449/UniFS

A-Evolve-Training: Autonomous Post-Training of a 30B Model

Authors: Zhan Shi, Bing He, Yisi Sang, Hanqing Lu, Benoit Dumoulin

Abstract: Post-training a frontier model is normally weeks of human work: proposing data and recipe changes, launching runs, reading evals, deciding what to keep. We report an autonomous system that runs this loop with no human in the loop, post-training a 30B Nemotron across four rounds over multiple weeks. The autonomously produced model reaches a held-out score of 0.86 against the top human submission's… ▽ More Post-training a frontier model is normally weeks of human work: proposing data and recipe changes, launching runs, reading evals, deciding what to keep. We report an autonomous system that runs this loop with no human in the loop, post-training a 30B Nemotron across four rounds over multiple weeks. The autonomously produced model reaches a held-out score of 0.86 against the top human submission's 0.87 on the public NVIDIA Nemotron-Reasoning Challenge leaderboard, placing 8th of ~4000 at the time of writing. More striking than the number: the loop detected that its own dev metric had stopped tracking external performance on the weakest domain -- candidates drove dev to record highs without moving the external target -- and revised its own search policy, no longer maximizing dev but seeking interventions that lowered the now-misleading proxy while improving the external target. We treat this as direct, auditable evidence that a scaled autonomous loop can produce discovery, not only optimization: it detected that its measurement frame had become misleading and changed what counted as evidence. We take the operational view that any system worth the "recursive self-improvement" label must eventually perform end-to-end post-training of a frontier-class model; this is one datapoint of that bar being cleared. We do not claim a "first autonomous match" of human researchers. The claim we make is narrower and auditable: to our knowledge, this is the first publicly reported autonomous post-training run at this scale, where prior public autonomous-ML-research demonstrations sit at GPT-2-class (~124M) budgets. The same system also post-trains the 120B and 550B Nemotron; with no public human baseline there, this shows only that the loop closes at that scale, not that its output is competitive -- infrastructure evidence, with the effectiveness claim deferred until a comparable human anchor exists. △ Less

Submitted 25 June, 2026; v1 submitted 9 June, 2026; originally announced June 2026.

Comments: 12 pages,

Approaching Shannon Bound with Lossless LLM Weight Compression

Authors: Hongshi Tan, Yao Chen, Gustavo Alonso, Weng-Fai Wong, Bingsheng He

Abstract: Large language models (LLMs) now scale to trillions of parameters, driving weight storage into the terabyte regime and creating an acute mismatch with GPU memory capacity. Although lossless compression is widely effective in other domains, it remains underutilized in LLM systems. Through a comprehensive entropy study across models from 1.5B to 405B parameters and numeric formats ranging from bf16… ▽ More Large language models (LLMs) now scale to trillions of parameters, driving weight storage into the terabyte regime and creating an acute mismatch with GPU memory capacity. Although lossless compression is widely effective in other domains, it remains underutilized in LLM systems. Through a comprehensive entropy study across models from 1.5B to 405B parameters and numeric formats ranging from bf16 to int4 and AWQ/SQ8, we find that LLM weights contain far less intrinsic randomness than their stored bitwidth implies, their effective entropy is 2-10x lower, indicating that up to a 10x footprint reduction is theoretically achievable without altering any weight values. Leveraging this insight, we introduce a tile-level, on-the-fly lossless decompression framework based on Asymmetric Numeral Systems that aligns decoding with the GEMM tiling pattern of GPU inference. Our design achieves bit-rates within 0.01-0.1 bits of the Shannon limit across a wide range of LLM numerical formats, demonstrating that nearly all statistical redundancy is eliminated. Integrated into the SGLang serving framework with multi-GPU support, our approach increases the maximum batch size of Qwen-14B from 47 to 75, improving throughput by up to 1.2x. On Mixtral-176B, the feasible batch size increases from 20 to 95 (4.8x), yielding up to 1.6x throughput improvement. Compared to state-of-the-art lossless compression approaches NeuZip and DFloat11, our design further improves throughput by up to 11x. △ Less

Submitted 14 June, 2026; originally announced June 2026.

Comments: Accepted to ISCA 2026

MA-ProofBench: A Two-Tiered Evaluation of LLMs for Theorem Proving in Mathematical Analysis

Authors: Lushi Pu, Weiming Zhang, Xinheng Xie, Zixuan Fu, Bingxiang He, Hongya Lyu, Xin Li, Jie Zhou, Yudong Wang

Abstract: Large Language Models (LLMs) have made notable progress in automated theorem proving, yet existing formal benchmarks remain limited in both mathematical coverage and difficulty. Most are concentrated in areas that are easier to formalize, such as algebra and elementary number theory, and provide limited coverage of subfields that require deeper reasoning, including mathematical analysis. To addres… ▽ More Large Language Models (LLMs) have made notable progress in automated theorem proving, yet existing formal benchmarks remain limited in both mathematical coverage and difficulty. Most are concentrated in areas that are easier to formalize, such as algebra and elementary number theory, and provide limited coverage of subfields that require deeper reasoning, including mathematical analysis. To address this gap, we introduce MA-ProofBench, to the best of our knowledge, the first formal theorem-proving benchmark dedicated to Mathematical Analysis. The benchmark contains 200 formalized theorems covering 6 core topics and 27 subcategories, including measure and integration theory, complex analysis, and functional analysis. The problems are divided into two difficulty levels, an undergraduate level (Level I, 100 problems) and a Ph.D. qualifying level (Level II, 100 problems), to evaluate how well LLMs perform formal reasoning at different mathematical depths. Each problem is constructed through a human-led, LLM-assisted formalization pipeline followed by independent expert review, ensuring that the formal statements remain faithful to the original mathematics. We evaluate a range of recent general-purpose reasoning models and formal theorem provers on MA-ProofBench. However, most models perform poorly: even the best-performing model, GPT-5.5, achieves only 16% Pass@8 on Level I and 5% on Level II, while most models stay close to 0% on Level II. Further analysis identifies Mathlib hallucinations and incomplete proofs as the two dominant failure modes, while an evaluation on the natural-language version of the benchmark exposes a clear gap between informal and formal reasoning. MA-ProofBench is intended to serve as a reliable reference for tracking progress in formal mathematical reasoning in advanced domains. △ Less

Submitted 15 June, 2026; v1 submitted 11 June, 2026; originally announced June 2026.

Comments: 19 pages, 4 figures, 4 tables

Verifiable Environments Are LEGO Bricks: Recursive Composition for Reasoning Generalization

Authors: Hao Xiang, Qiaoyu Tang, Le Yu, Yaojie Lu, Xianpei Han, Ben He, Le Sun, Bowen Yu, Peng Wang, Hongyu Lin, Dayiheng Liu

Abstract: Reinforcement Learning (RL) with verifiable environments has emerged as a powerful approach for enhancing the reasoning capabilities of Large Language Models (LLMs). While prior research demonstrates that scaling environment quantity improves RL performance, existing manual or individual construction methods suffer from linear scaling limits, thereby hindering scalable reasoning generalization. Th… ▽ More Reinforcement Learning (RL) with verifiable environments has emerged as a powerful approach for enhancing the reasoning capabilities of Large Language Models (LLMs). While prior research demonstrates that scaling environment quantity improves RL performance, existing manual or individual construction methods suffer from linear scaling limits, thereby hindering scalable reasoning generalization. This paper introduces RACES (\textbf{R}ecursive \textbf{A}utomated \textbf{C}omposition for \textbf{E}nvironment \textbf{S}caling), a framework that conceptualizes verifiable environments as composable building blocks that can be recursively assembled. The key insight is that when the codomain (output type) of one environment matches the domain (input type) of another, they can be automatically fused into a new verifiable environment, enabling recursive composition. RACES is implemented with 300 individual environments and defines a set of composition operators (\textsc{SEQUENTIAL}, \textsc{PARALLEL}, \textsc{SORT}, and \textsc{SELECT}) that induce diverse reasoning patterns. Extensive experiments show that RL training on these composite environments consistently enhances reasoning generalization. Specifically, RACES improves DeepSeek-R1-Distill-Qwen-14B by an average of 3.1 points (from 48.2 to 51.3) and boosts Qwen3-14B performance from 58.8 to 61.1 on six benchmarks, which are unseen during the construction of training environments. Moreover, RACES achieves performance comparable to training on 300 individual environments using only 50 base environments, demonstrating significant efficiency in environment utilization. △ Less

Submitted 10 June, 2026; originally announced June 2026.

IDEAL: In-DEpth ALignment Makes A Discrete Representation AutoEncoder

Authors: Yitong Chen, Zijie Diao, Junke Wang, Lingyu Kong, Yixuan Ren, Bo He, Yu-Gang Jiang, Zuxuan Wu

Abstract: Built on pretrained vision foundation models (VFMs), representation autoencoders (RAEs) have recently emerged as a promising approach for constructing semantically rich latent spaces for image generation. However, their reconstruction quality often remains suboptimal, largely because deep VFM representations do not preserve sufficient fine-grained visual detail. This limitation becomes even more s… ▽ More Built on pretrained vision foundation models (VFMs), representation autoencoders (RAEs) have recently emerged as a promising approach for constructing semantically rich latent spaces for image generation. However, their reconstruction quality often remains suboptimal, largely because deep VFM representations do not preserve sufficient fine-grained visual detail. This limitation becomes even more severe after discretization, where missing low-level information is difficult to recover. In fact, we observe that shallow VFM features retain considerably richer local appearance and structural detail, which complements the high-level semantics carried by deep features used in existing RAEs. Motivated by this complementary property, we propose Ideal, an In-depth Alignment framework for discrete representation autoencoding. By jointly aligning quantized tokens with both shallow and deep VFM features, Ideal enables the resulting discrete visual tokens to preserve both visual fidelity and rich semantics. Extensive experiments demonstrate that Ideal yields superior reconstruction performance, achieving 0.61 rFID on ImageNet and outperforming the previous best method by 0.28. When used for autoregressive image generation, Ideal further produces a gFID of 1.89, establishing a new state of the art for autoregressive image generation. △ Less

Submitted 9 June, 2026; originally announced June 2026.

Comments: Code is available at https://github.com/Row11n/IDEAL

Improving Text-Instance Alignment Of Foreground Conditioned Out-Painting Via Customized Concept Embedding

Authors: Yihao Zhao, Xuan Han, Bin He, Mingyu You

Abstract: To showcase products, merchants often incur substantial costs creating high-quality display images. Foreground Conditioned Outpainting (FCO) meets this demand, allowing users to create desired backgrounds for foreground instances at a low cost by adjusting the text prompt. However, existing text-driven FCO methods exhibit critical flaws in their outputs, most notably the presence of artifacts, whi… ▽ More To showcase products, merchants often incur substantial costs creating high-quality display images. Foreground Conditioned Outpainting (FCO) meets this demand, allowing users to create desired backgrounds for foreground instances at a low cost by adjusting the text prompt. However, existing text-driven FCO methods exhibit critical flaws in their outputs, most notably the presence of artifacts, which refer to regions in the synthesized background that share the same semantics as the foreground instance. Such artifacts diminish the object's prominence and degrade image quality. We attribute the issue to the misalignment between the given instance and text-derived concept embeddings. To address this, we propose the Customized Concept Embedding Diffusion (CCE-Diffusion) framework. Its core is a CCE-Module to customize concept embeddings, bridging the gap between generic noun semantics and a specific visual instance. An Instance-Aware Loss guides the module's optimization, while a Semantic-Preserving Prompt Template prevents customized embeddings from distorting other words in the prompt. Both qualitative and quantitative evaluations demonstrate that CCE-Diffusion significantly reduces artifacts in the outputs. As a plug-and-play component, the CCE-Module can integrate with various FCO methods, enhancing their performance. △ Less

Submitted 9 June, 2026; originally announced June 2026.

Hardware-accelerated Aggregation: Unification and Specialization

Authors: Alireza Shateri, Hongshi Tan, Michael Ng, Bingsheng He, Qizhen Zhang

Abstract: The high efficiency of domain-specific hardware has sparked substantial interest in adopting accelerators in data analytics systems. Among many choices, GPUs and FPGAs thrived as two popular solutions due to their prevalent deployments in cloud data centers. This paper investigates hardware acceleration solutions for aggregation, a critical data analytics operation. Specifically, we implement aggr… ▽ More The high efficiency of domain-specific hardware has sparked substantial interest in adopting accelerators in data analytics systems. Among many choices, GPUs and FPGAs thrived as two popular solutions due to their prevalent deployments in cloud data centers. This paper investigates hardware acceleration solutions for aggregation, a critical data analytics operation. Specifically, we implement aggregation with a unified hardware acceleration framework, which trades efficiency for ease of programming and portability, and then further develop hardware-specific optimizations. We evaluate these solutions on three recent computing hardware platforms: a CPU, a GPU, and an FPGA, with metrics that cover both the performance and energy consumption of on-device and end-to-end processing. △ Less

Submitted 8 June, 2026; originally announced June 2026.

ACM Class: H.2.2; B.2.4

AdaPlanBench: Evaluating Adaptive Planning in Large Language Model Agents under World and User Constraints

Authors: Jiayu Liu, Cheng Qian, Zhenhailong Wang, Bingxuan Li, Jiateng Liu, Qing Zong, Heng Wang, Jeonghwan Kim, Yumeng Wang, Bingxiang He, Xiusi Chen, Yi R. Fung, Heng Ji

Abstract: Planning for real-world problems by language models often involves both world and user constraints, which may not be fully specified upfront and are progressively disclosed through interaction. However, existing benchmarks still underexplore adaptive planning under such progressively revealed dual constraints. To address this gap, we introduce AdaPlanBench, a dynamic interactive benchmark for eval… ▽ More Planning for real-world problems by language models often involves both world and user constraints, which may not be fully specified upfront and are progressively disclosed through interaction. However, existing benchmarks still underexplore adaptive planning under such progressively revealed dual constraints. To address this gap, we introduce AdaPlanBench, a dynamic interactive benchmark for evaluating whether Large Language Model (LLM) agents can adaptively plan and re-plan under progressively revealed world and user constraints. AdaPlanBench is built on 307 household tasks, with a scalable constraint construction pipeline that augments each task with dual constraints. At runtime, agents interact with the environment in a multi-turn protocol where hidden constraints are revealed only when the agent proposes a plan that violates them, requiring iterative plan revision under accumulating feedback. This makes planning challenging, as agents must infer and track constraints from feedback while re-planning effectively. Experiments on ten leading LLMs show that adaptive planning under dual constraints remains challenging, with the best model reaching only 67.75% accuracy. We further observe that performance degrades as more constraints accumulate, with user constraints posing a particularly large challenge and failures often stemming from weaker physical grounding and reduced effectiveness. These results establish AdaPlanBench as a testbed for dual-constrained interactive planning and highlight the challenge of reliable adaptation to dynamically revealed constraints in LLM agents. △ Less

Submitted 9 July, 2026; v1 submitted 3 June, 2026; originally announced June 2026.

Comments: COLM 2026