Showing 1–50 of 290 results for author: Cui, Q

GateDiffInt: Gate-Mediated Controllable Diffusion and Multi-Intent LLM Distillation for User Behavior Modeling

Authors: Jialong Duan, Zichen Zhang, Zirui Tu, Zheng Zhang, Zepeng Li, Qingyao Cui, Qinwen Wang, Yudan Liu, Luo Yang, Yao Hu

Abstract: Existing ranking models encode intent only implicitly, making it hard to disentangle structured intents of varying strength and temporal scale. Noise and intent in behavior sequences are mutually reinforcing---we call this Noise--Intent Coupling (NIC). Noise dilutes true intents, while the lack of structured intent priors leaves denoising without a clear target. To address NIC, we propose GateDiff… ▽ More Existing ranking models encode intent only implicitly, making it hard to disentangle structured intents of varying strength and temporal scale. Noise and intent in behavior sequences are mutually reinforcing---we call this Noise--Intent Coupling (NIC). Noise dilutes true intents, while the lack of structured intent priors leaves denoising without a clear target. To address NIC, we propose GateDiffInt, an intent interaction framework for industrial ranking. It uses the final conversion signal to jointly align sequence denoising and intent extraction. GateDiffInt applies a controllable forward diffusion process with dual gating to enhance and denoise behavior sequences. A large language model then acts as teacher to distill four structured intents---long-term, short-term, latent, and conversion into a lightweight student model. The enhanced sequence and structured intent representations are deeply fused via attention to produce intent-aware representations for conversion-rate prediction. Extensive experiments on public and large-scale industrial datasets show consistent gains over strong baselines. In online A/B tests serving hundreds of millions of daily active users, GateDiffInt delivers substantial GMV improvements and has been deployed to primary traffic, confirming both effectiveness and production readiness. △ Less

Submitted 19 August, 2026; v1 submitted 19 August, 2026; originally announced August 2026.

CTIFoundry: An Agent-Native Corpus Scaffold for Cyber Threat Intelligence

Authors: Yutong Cheng, Changze Li, Qian Cui, Wei Ding, Lingzhi Wang, Yan Chen, Peng Gao

Abstract: Cyber threat intelligence (CTI) is increasingly consumed not by human analysts but by LLM agents that compose multi-step investigations at query time. The harness side of this shift has matured rapidly (planning loops, tool protocols, context management), but the corpus side has not: threat reports and vulnerability databases are still packaged for retrieval-augmented generation, as opaque chunks… ▽ More Cyber threat intelligence (CTI) is increasingly consumed not by human analysts but by LLM agents that compose multi-step investigations at query time. The harness side of this shift has matured rapidly (planning loops, tool protocols, context management), but the corpus side has not: threat reports and vulnerability databases are still packaged for retrieval-augmented generation, as opaque chunks behind an embedding index. We argue that this substrate, not model capability, is the bottleneck on agentic CTI investigation, and present CTIFoundry, an agent-native corpus scaffold. At build time, CTIFoundry materializes the latent structure of a CTI corpus: a deterministic ontology graph over four authoritative knowledge bases (CVE, CWE, CAPEC, ATT&CK) whose official cross-references become typed, traversable edges; a span-grounded report layer whose canonical, alias-resolved cross-vendor entities index provenance-carrying chunks; and hybrid dense+lexical retrieval surfaces. At query time this structure is exposed through seven typed tools and three procedural skills mounted on a stock open-source agent harness. On the public CTIConnect benchmark, swapping only the action surface lifts the identically-harnessed agent by +0.19 to +0.28 overall F1 across a four-model, two-provider panel: a small model on CTIFoundry surpasses a flagship on the flat substrate, and the gain is not bought with search effort, since on both Claude models the scaffolded agent is more accurate at roughly half the tool calls. An ablation attributes it: typed structure carries the larger share, procedural skills convert structure into discipline, and the two compose super-additively, because skills bind only to structure that exists. △ Less

Submitted 19 August, 2026; originally announced August 2026.

Comments: Preprint

On the saturation number of the kite graph

Authors: Huanying Bian, Qing Cui, Shengjin Ji, Fufong Ma

Abstract: For a fixed graph $H$, a graph $G$ is $H$-saturated if $G$ does not contain a copy of $H$, but adding any edge $e \in E(\overline{G})$ to $G$ creates a copy of $H$. The saturation number $\mathrm{sat}(n,H)$ is the minimum number of edges in an $H$-saturated graph on $n$ vertices. Let $K$ be the kite graph, formed by removing one edge from $ K_4$ and then attaching a pendant edge to a vertex of deg… ▽ More For a fixed graph $H$, a graph $G$ is $H$-saturated if $G$ does not contain a copy of $H$, but adding any edge $e \in E(\overline{G})$ to $G$ creates a copy of $H$. The saturation number $\mathrm{sat}(n,H)$ is the minimum number of edges in an $H$-saturated graph on $n$ vertices. Let $K$ be the kite graph, formed by removing one edge from $ K_4$ and then attaching a pendant edge to a vertex of degree two in the resulting graph.In this paper, we first establish a relationship between connectivity and $K$-saturated graphs, and subsequently determine the saturation number of the kite graph $K$. Moreover, we completely characterize all extremal graphs.Our result provides a partial answer to a problem raised by Hua and Peng [Discrete Math. 349 (2026) 114674]. △ Less

Submitted 17 August, 2026; originally announced August 2026.

Comments: 18 pages, 9 figures

Mapping the City Through the Lens of Language Models

Authors: Wanqi Liu, Rong Zhao, Zhizhou Sha, Qinyu Cui, Yecheng Zhang

Abstract: Language models often complete an underspecified reference to a city with unstated assumptions about urban size, form, infrastructure, environment, and function. We measure those assumptions without naming places. Ten open-weight checkpoints rate anonymized profiles derived from real morphological urban centres across 40 audited indicators and seven domains. The design combines constrained probabi… ▽ More Language models often complete an underspecified reference to a city with unstated assumptions about urban size, form, infrastructure, environment, and function. We measure those assumptions without naming places. Ten open-weight checkpoints rate anonymized profiles derived from real morphological urban centres across 40 audited indicators and seven domains. The design combines constrained probability-based ratings, prespecified reliability screens, lineage-aware aggregation, multiple population weightings, an independent replication sample, and whole-profile validation. The clearest shared tendency favours urban profiles with larger developed area, faster recent growth, greater mapped infrastructure and non-residential capacity, and less sparse form. Most eligible directions recur in the replication data, and direct ratings of complete profiles show moderate agreement with the indicator-wise construction. Geographic differences shrink after accounting for city scale and development, while reliably measured paired tasks indicate that typicality and desirability are often closely aligned. The framework makes an otherwise vague notion of what models regard as an ordinary city empirically traceable. The resulting evidence delineates a shared yet model-dependent portrait of the city through the lens of language models. △ Less

Submitted 7 August, 2026; v1 submitted 3 August, 2026; originally announced August 2026.

LC-SEPLM: long-range contact-supervised adaptation for sequence-only protein representation learning

Authors: Chen Wang, Boming Kang, Qinghua Cui

Abstract: Protein language models learn transferable sequence representations. However, because they primarily model contextual dependencies along amino-acid sequences, their training objectives do not explicitly constrain the model to learn three-dimensional residue contacts formed after folding . Here, we introduce LC-SEPLM (Long-range Contact-supervised ESM Protein Language Model), which adapts ESM2 with… ▽ More Protein language models learn transferable sequence representations. However, because they primarily model contextual dependencies along amino-acid sequences, their training objectives do not explicitly constrain the model to learn three-dimensional residue contacts formed after folding . Here, we introduce LC-SEPLM (Long-range Contact-supervised ESM Protein Language Model), which adapts ESM2 with LoRA and long-range residue-pair contact supervision while retaining sequence-only downstream inference. Pair-specific queries use cross-attention over the complete sequence to extract global sequence context associated with long-range spatial contacts. To expose the model to diverse structural information, we trained LC-SEPLM on 500,000 AlphaFold Swiss-Prot proteins. In downstream evaluation, LC-SEPLM improved all eight protein-level tasks relative to ESM2. The largest gain occurred in remote-homology recognition, where macro-F1 increased from 0.6122 to 0.6769 (+0.0647, or 6.47 percentage points). On the official ESM-S EC benchmark, LC-SEPLM also outperformed ESM-S with a maximum absolute gain of 0.1771. These results support residue-pair contact supervision as a bounded route for introducing structural information into protein sequence representations while preserving sequence-only inference. △ Less

Submitted 28 July, 2026; v1 submitted 24 July, 2026; originally announced July 2026.

A curvature characterization of the Cartan minimal hypersurface in $\mathbb S^5$

Authors: Qing Cui

Abstract: Lawson showed that a non-totally geodesic Einstein minimal hypersurface in $\mathbb S^5$ is congruent to the Clifford hypersurface $\mathbb S^2(1/\sqrt2)\times \mathbb S^2(1/\sqrt2).$ It is also known, by work of Cartan and Ôtsuki, that a non-totally geodesic locally conformally flat minimal hypersurface in $\mathbb S^5$ is of Ôtsuki type, including the Clifford hypersurface… ▽ More Lawson showed that a non-totally geodesic Einstein minimal hypersurface in $\mathbb S^5$ is congruent to the Clifford hypersurface $\mathbb S^2(1/\sqrt2)\times \mathbb S^2(1/\sqrt2).$ It is also known, by work of Cartan and Ôtsuki, that a non-totally geodesic locally conformally flat minimal hypersurface in $\mathbb S^5$ is of Ôtsuki type, including the Clifford hypersurface $\mathbb S^1(1/2)\times \mathbb S^3(\sqrt3/2).$ In this paper we study closed minimal hypersurfaces $M$ in $\mathbb S^5$ satisfying $|W|^2=2|\mathring{\operatorname{Ric}}|^2,$ where $W$ is the Weyl tensor and $\mathring{\operatorname{Ric}}$ is the trace-free Ricci tensor. We call this the Euler-balanced condition. We prove that such a hypersurface is either totally geodesic or congruent to the Cartan minimal hypersurface. △ Less

Submitted 12 July, 2026; originally announced July 2026.

Comments: 22 pages, no figure. Comments are welcome

Towards CSI-Native Foundation Models: A Channel-Adaptive Roadmap for 6G

Authors: Chenyu Zhang, Xinchen Lyu, Chenshan Ren, Shuhan Liu, Qimei Cui

Abstract: Wireless foundation models offer a path toward reusable channel state information (CSI) intelligence for sixth-generation (6G) systems. However, existing generic-backbone adaptation and CSI pretraining methods often treat CSI as task tensors rather than propagation-conditioned channel responses, thereby failing to capture the intrinsic time-frequency-spatial geometry of wireless environments. This… ▽ More Wireless foundation models offer a path toward reusable channel state information (CSI) intelligence for sixth-generation (6G) systems. However, existing generic-backbone adaptation and CSI pretraining methods often treat CSI as task tensors rather than propagation-conditioned channel responses, thereby failing to capture the intrinsic time-frequency-spatial geometry of wireless environments. This paper presents a channel-adaptive roadmap toward CSI-native foundation models, proposing a unified framework that aligns pretraining, positional modeling, and attention control with three channel requirements: scale-aware heterogeneous exposure, physical time-frequency-antenna coordinates, and correlation-bounded token interaction. Extensive experiments demonstrate the superiority of the proposed framework across three dimensions: zero-shot generalization, reducing NMSE by more than 4 dB across spatial-temporal-frequency tasks; scale extrapolation, yielding up to a 5.4 dB gain under 8 times unseen antenna scaling; and inference efficiency, accelerating mobility-aware processing by up to 18.8%. A system-level evaluation with Sionna SYS further shows that the proposed framework uses only 7.01% of dense-pilot overhead, reaches -18.64 dB average NMSE, and improves average net spectral efficiency by 36.6% over dense LMMSE and 15.5% over WiFo, indicating that CSI-native representation learning can support pilot-efficient radio access. △ Less

Submitted 12 June, 2026; originally announced June 2026.

Comments: 7 pages, 5 figures, submmited to IEEE WCM

Ling and Ring 2.6 Technical Report: Efficient and Instant Agentic Intelligence at Trillion-Parameter Scale

Authors: Ang Li, Ben Liu, Bin Han, Bin Hu, Bin Jing, Binbin Hu, Bing Li, Cai Chen, Caizhi Tang, Changxin Tian, Chao Huang, Chao Zhang, Chen Liang, Chen Qian, Chengfu Tang, Chengyao Wen, Chilin Fu, Chunwei Wu, Cong Zhang, Cunyin Peng, Daixin Wang, Dalong Zhang, Deng Zhao, Dingnan Jin, Dingyuan Zhu , et al. (193 additional authors not shown)

Abstract: Efficient and scalable agentic intelligence requires models that can deliver both low-latency responses and strong reasoning capabilities while remaining practical to train, serve, and deploy. In this report, we present Ling-2.6 and Ring-2.6, a family of models designed to address this challenge at scale. Ling-2.6 is optimized for instant response generation and high capability per output token, w… ▽ More Efficient and scalable agentic intelligence requires models that can deliver both low-latency responses and strong reasoning capabilities while remaining practical to train, serve, and deploy. In this report, we present Ling-2.6 and Ring-2.6, a family of models designed to address this challenge at scale. Ling-2.6 is optimized for instant response generation and high capability per output token, whereas Ring-2.6 is tailored for deeper reasoning and more advanced agentic workflows. Instead of training from scratch, we upgrade the Ling-2.0 base model through architectural migration pre-training and large-scale post-training. This upgrade is guided by a unified co-design of model architecture, optimization objectives, serving systems, and agent training environments, enabling improvements in both model capability and deployment efficiency. At the architectural level, we introduce a hybrid linear attention design that integrates Lightning Attention with MLA, improving the efficiency of long-context training and decoding. To further enhance token efficiency, we optimize capability per output token through Evolutionary Chain-of-Thought, Linguistic Unit Policy Optimization, bidirectional preference alignment, and shortest-correct-response distillation. For agentic capabilities, we propose KPop, a reinforcement learning framework designed to support stable training of Ring-2.6-1T on large-scale environment-grounded data. KPop improves training efficiency through asynchronous scheduling across coding, search, tool use, and workflow execution, enabling scalable learning from complex agent-environment interactions. Together, Ling-2.6 and Ring-2.6 provide a practical pathway toward efficient, scalable, and open agentic systems. We open-source all checkpoints in the 2.6 family to support further research and development in practical agentic intelligence. △ Less

Submitted 12 June, 2026; originally announced June 2026.

Improving Cross-Format Robustness in Language Models with Multi-Format Training

Authors: June M. Liu, Shaomian Zheng, He Cao, Dingnan Jin, Qing Cui, Jun Zhou

Abstract: Large language models often remain sensitive to answer format: a question solved correctly in one form may fail in another semantically equivalent form. To study this gap, we define cross-format robustness as the extent to which a model answers the same underlying question consistently across formats. We then compare full-format training with FormatMix, which expands only a subset of training item… ▽ More Large language models often remain sensitive to answer format: a question solved correctly in one form may fail in another semantically equivalent form. To study this gap, we define cross-format robustness as the extent to which a model answers the same underlying question consistently across formats. We then compare full-format training with FormatMix, which expands only a subset of training items into multiple equivalent formats using either random or targeted selection. Across GLM4 and Llama-3.1, multi-format supervision consistently improves both task performance and cross-format robustness, whereas Multiple-choice question (MCQ)-only supervision alone brings little benefit and can even reduce robustness. We further find that expanding only about 30% of the training set into multiple formats often recovers most of the gain from full-format training, and this effect appears across the model families and sizes we study. These results suggest that format diversity, rather than additional supervision alone, is the key driver of robustness. That lightweight multi-format augmentation is a practical way to make LLMs less sensitive to answer format without changing the base model. △ Less

Submitted 10 June, 2026; originally announced June 2026.

Ferroelectrical Switching as a Probe of Quantum Damping in Magnetic Spin Systems

Authors: Yuefei Liu, Anna Delin, Olle Eriksson, Erik Sjöqvist, Kaiyou Wang, Qirui Cui

Abstract: While damped spin dynamics is important for the understanding of magnetic materials, clear signatures of \emph{quantum corrections} to the Gilbert damping mechanism remain elusive. We propose a route to distinguish quantum and classical Gilbert spin damping using ferroelectric control of a magnetic dimer. Ab initio calculations for dimers on ferroelectric substrates show that polarization reversal… ▽ More While damped spin dynamics is important for the understanding of magnetic materials, clear signatures of \emph{quantum corrections} to the Gilbert damping mechanism remain elusive. We propose a route to distinguish quantum and classical Gilbert spin damping using ferroelectric control of a magnetic dimer. Ab initio calculations for dimers on ferroelectric substrates show that polarization reversal switches the inter-spin exchange between ferromagnetic and antiferromagnetic regimes. We formulate a magnetization-based diagnostic that relates magnetization traces to entanglement dynamics, which enables ferroelectrical on/off control of dimer entanglement. Material-informed quantum Landau-Lifshitz-Gilbert simulations illustrate how the signature of magnetization dynamics can, in principle, be used to infer the existence of quantum Gilbert spin damping. This minimal and non-volatile platform connects first-principles modeling to experimentally accessible observables and provides a starting point for voltage-controlled quantum entanglement in magnetic spin networks. △ Less

Submitted 5 June, 2026; originally announced June 2026.

Comments: 7 pages, 3 figures

A strict upper volume bound for minimal graphs in the unit ball

Authors: Qing Cui

Abstract: Let $u$ be a solution of the minimal surface equation on a domain containing the closed unit ball $\overline{B^n}\subset\mathbb R^n$. A classical calibration argument gives $|Graph_u\cap B^{n+1}| \leq \frac12 |\mathbb S^n|.$ A basic question is whether this half-sphere bound is sharp for minimal graphs. We show that it is not. More precisely, for every $n\geq2$, there exists an explicit constant… ▽ More Let $u$ be a solution of the minimal surface equation on a domain containing the closed unit ball $\overline{B^n}\subset\mathbb R^n$. A classical calibration argument gives $|Graph_u\cap B^{n+1}| \leq \frac12 |\mathbb S^n|.$ A basic question is whether this half-sphere bound is sharp for minimal graphs. We show that it is not. More precisely, for every $n\geq2$, there exists an explicit constant $δ_n>0$, depending only on $n$, such that $ | Graph_u\cap B^{n+1}|\leq \frac12 |\mathbb S^n|-δ_n$. The proof combines calibration with a canonical spherical filling associated with the subgraph and a quantitative incompatibility between near equality in the calibration estimate and the divergence-free structure of the minimal surface equation. We also give an improved explicit gap and formulate the corresponding sharp extremal problem. △ Less

Submitted 9 August, 2026; v1 submitted 31 May, 2026; originally announced June 2026.

Comments: 11 pages, no figure. This version replaces the withdrawn previous version. The previous sharpness claim was incorrect. The present version proves instead a uniform positive gap below the half-sphere bound. All comments are welcome!

Data-Driven Discovery of Unconventional Antiferromagnets

Authors: Qirui Cui, Chenxu Liu, Anna Delin, Kaiyou Wang

Abstract: Unconventional antiferromagnets combine zero net magnetization with spin-split electronic bands, offering a distinct, important platform for spintronics. Their discovery, however, has so far depended largely on case-by-case studies and on a limited number of compounds with experimentally resolved magnetic structures. Here, we overcome these bottlenecks by resolving magnetic ground states across a… ▽ More Unconventional antiferromagnets combine zero net magnetization with spin-split electronic bands, offering a distinct, important platform for spintronics. Their discovery, however, has so far depended largely on case-by-case studies and on a limited number of compounds with experimentally resolved magnetic structures. Here, we overcome these bottlenecks by resolving magnetic ground states across a broad materials database. We narrow down 37163 magnets from the Materials Project to 189 collinear antiferromagnets by combining physics-informed prescreening, high-throughput exchange calculations and Luttinger-Tisza analysis. Among these, symmetry analysis identifies 36 altermagnets and 11 Luttinger-compensated ferrimagnets (LCFs), including 22 altermagnets and 9 LCFs that have not been reported previously. The identified unconventional antiferromagnets can support nonrelativistic spin Hall effects and doping-tunable spin transport with switchable polarization and giant anisotropy. Our framework converts broad structural databases into a curated, symmetry-classified set of experimentally testable compensated spin-split magnets, establishing a scalable route for the efficient discovery of functional antiferromagnets. △ Less

Submitted 29 May, 2026; originally announced June 2026.

Comments: 22 pages, 3 figures

Skill is Not One-Size-Fits-All: Model-Aware Skill Alignment for LLM Agents

Authors: Jianxiang Yu, Jiapeng Zhu, Bochen Lin, Qier Cui, Zichen Ding, Xiang Li

Abstract: LLM agents increasingly retrieve externally curated skills-procedural instructions retrieved at decision time-to improve performance on long-horizon interactive tasks. Existing skill libraries are typically treated as model-agnostic, reusing the same skill formulations across backbones with substantially different capacities and behaviors. However, our controlled experiments across multiple model… ▽ More LLM agents increasingly retrieve externally curated skills-procedural instructions retrieved at decision time-to improve performance on long-horizon interactive tasks. Existing skill libraries are typically treated as model-agnostic, reusing the same skill formulations across backbones with substantially different capacities and behaviors. However, our controlled experiments across multiple model scales show that skill effectiveness is strongly model-dependent: a skill that benefits one backbone can harm another. Motivated by this observation, we propose MASA Model-Aware Skill Alignment, a framework that adapts skills to each target backbone without modifying agent weights. MASA operates in two stages: (1) a hierarchical skill evolution pipeline that iteratively rewrites general and task-specific skills using hill climbing and UCB-driven tree search, guided by environment feedback and model capability profiles; and (2) a lightweight model-conditioned skill rewriter trained on evolution trajectories to reproduce the adaptation in a single forward pass. Experiments across three interactive environments and four backbones show that MASA consistently achieves the best overall performance, with gains of up to 25.8 points over the strongest baseline. The learned rewriter further generalizes to unseen tasks and environments without additional search, consistently outperforming a much larger teacher LLM at a fraction of the inference cost. △ Less

Submitted 28 May, 2026; originally announced May 2026.

TTPrint: Evidence-Grounded TTP Extraction via Diverge-then-Converge Verification

Authors: Yutong Cheng, Changze Li, Raihan Sultan Pasha Basuki, Qian Cui, Wei Ding, Peng Gao

Abstract: Extracting MITRE ATT&CK techniques from cyber threat intelligence (CTI) reports is an open-set, multi-label problem requiring both high recall (not missing techniques) and high precision (not hallucinating unsupported ones). Existing methods--rule-based, supervised, and LLM-based--struggle to achieve both: rule-based and supervised approaches lack generalizability across diverse attack description… ▽ More Extracting MITRE ATT&CK techniques from cyber threat intelligence (CTI) reports is an open-set, multi-label problem requiring both high recall (not missing techniques) and high precision (not hallucinating unsupported ones). Existing methods--rule-based, supervised, and LLM-based--struggle to achieve both: rule-based and supervised approaches lack generalizability across diverse attack descriptions, while LLM-based approaches that couple candidate generation and validation within a single inference step suffer from limited recall and precision simultaneously. We propose TTPrint, which addresses this challenge through a diverge-then-converge design inspired by how human analysts work: first extracting broadly, then verifying rigorously. In the divergent phase, reports are decomposed into atomic behaviors and candidate techniques are proposed broadly. A deterministic span localization stage then anchors each candidate to a specific evidence window in the source text. A convergent verification stage retains only candidates supported by both the localized evidence and the authoritative MITRE definition. We contribute two evaluation resources--a cleaned TRAM benchmark (TRAM-Clean) and a new annotated dataset (TTPrint-Bench)--to address known annotation noise in existing benchmarks and elevate the task to document-level TTP extraction. On TRAM-Clean and TTPrint-Bench, TTPrint achieves 76.48% and 87.39% macro-F1 respectively, outperforming the leading baseline by 63.5% and 29.4%. A multi-backbone analysis across six LLMs and a threshold sensitivity study further demonstrate generalizability across model choices and provide practical guidance for parameter selection. △ Less

Submitted 25 May, 2026; originally announced May 2026.

Comments: Preprint

What Really Improves Mathematical Reasoning: Structured Reasoning Signals Beyond Pure Code

Authors: Yuze Zhao, Junpeng Fang, Lu Yu, Zhenya Huang, Kai Zhang, Qing Cui, Qi Liu, Jun Zhou, Enhong Chen

Abstract: Code has become a standard component of modern foundation language model (LM) training, yet its role beyond programming remains unclear. We revisit the claim that code improves reasoning through controlled pretraining experiments on a 10T-token corpus with fine-grained domain separation. Our findings are threefold. First, when code is restricted to standalone executable programs and Code-NL data a… ▽ More Code has become a standard component of modern foundation language model (LM) training, yet its role beyond programming remains unclear. We revisit the claim that code improves reasoning through controlled pretraining experiments on a 10T-token corpus with fine-grained domain separation. Our findings are threefold. First, when code is restricted to standalone executable programs and Code-NL data are controlled for, code substantially improves programming ability but does not act as a general reasoning enhancer; instead, it competes with knowledge-intensive tasks, especially complex mathematical reasoning. Second, the reasoning gains often attributed to code are better explained by cross-domain structured reasoning traces, such as code-text and math-text mixtures, rather than by executable code alone. Third, increasing the density of structured math-domain samples within a fixed math budget yields substantial gains on difficult mathematical reasoning while largely preserving programming performance, suggesting that cognitive scaffolds offer a targeted way to mitigate cross-domain trade-offs. Finally, routing analyses show that data-composition effects are reflected in expert-activation patterns, providing mechanism-level evidence for competitive and synergistic interactions across domains. Our results clarify which data characteristics transfer across capability dimensions and point to more precise data-centric optimization strategies. △ Less

Submitted 19 May, 2026; originally announced May 2026.

Comments: Accepted by ICML 2026, 22 pages, 10 figures

3D Primitives are a Spatial Language for VLMs

Authors: Junze Liu, Kun Qian, Florian Dubost, Kai Zhong, Arvind Srinivasan, Nan Chen, Anping Wang, Sam Zhang, Alejandro Mottini, Qingjun Cui, Tian Wang

Abstract: Vision-language models (VLMs) exhibit a striking paradox: they can generate executable code that reconstructs a 3D scene from geometric primitives with correct object counts, classes, and approximate positions, yet the same models fail at simpler spatial questions on the same image. We show that 3D geometric primitives (cubes, spheres, cylinders, expressed in executable code) serve as a powerful i… ▽ More Vision-language models (VLMs) exhibit a striking paradox: they can generate executable code that reconstructs a 3D scene from geometric primitives with correct object counts, classes, and approximate positions, yet the same models fail at simpler spatial questions on the same image. We show that 3D geometric primitives (cubes, spheres, cylinders, expressed in executable code) serve as a powerful intermediate representation for spatial understanding, and exploit this through three contributions. First, we introduce \textbf{\textsc{SpatialBabel}}, a benchmark evaluating fourteen VLMs on primitive-based 3D scene reconstruction across six \emph{scene-code languages} (programming languages and declarative formats for 3D primitive scenes), revealing that a single model's object-detection F1 can vary by up to $5.7\times$ across languages. Second, we propose \textbf{Code-CoT} (Code Chain-of-Thought), a training-free inference strategy that routes spatial reasoning through primitive-based code generation. Code-CoT lifts the SpatialBabel-QA-Score by up to $+6.4$\% on primitive scenes and real-photo CV-Bench-3D accuracy by $+5.0$\% for VLMs with strong coding capabilities. Third, we propose \textbf{S$^{3}$-FT} (Self-Supervised Spatial Fine-Tuning), which self-supervisedly distills primitive spatial knowledge into general visual reasoning by parsing the model's own Three.js primitive-reconstructions into structured annotations and fine-tuning on the result, with \emph{no human labels and no teacher model}. Training on primitive images alone, S$^3$-FT improves Qwen3-VL-8B by $+4.6$ to $+8.6$\% on SpatialBabel-Primitive-QA, $+9.7$\% on CV-Bench-2D, and $+17$\% on HallusionBench; the recipe transfers across model families. These results establish geometric primitives in code as both a diagnostic and a transferable spatial vocabulary for VLMs. We will release all artifacts upon publication. △ Less

Submitted 12 May, 2026; originally announced May 2026.

DiffScore: Text Evaluation Beyond Autoregressive Likelihood

Authors: Wen Lai, Yingli Shen, Dingnan Jin, Qing Cui, Jun Zhou, Maosong Sun, Alexander Fraser

Abstract: Autoregressive language models are widely used for text evaluation, however, their left-to-right factorization introduces positional bias, i.e., early tokens are scored with only leftward context, conflating architectural asymmetry with true text quality. We propose masked reconstruction as an alternative paradigm, where every token is scored using full bidirectional context. We introduce DiffScor… ▽ More Autoregressive language models are widely used for text evaluation, however, their left-to-right factorization introduces positional bias, i.e., early tokens are scored with only leftward context, conflating architectural asymmetry with true text quality. We propose masked reconstruction as an alternative paradigm, where every token is scored using full bidirectional context. We introduce DiffScore, an evaluation framework built on Masked Large Diffusion Language Models. By measuring text recoverability across continuous masking rates, DiffScore eliminates positional bias and naturally establishes an evaluation hierarchy from local fluency to global coherence. We further provide diagnostic tools unavailable to autoregressive frameworks: multi-timestep quality profiles that decompose scores across masking rates, and bidirectional PMI decomposition that disentangles fluency from faithfulness. Experiments across ten benchmarks show that DiffScore consistently outperforms autoregressive baselines in both zero-shot and fine-tuned settings. The code is released at: https://github.com/wenlai-lavine/DiffScore. △ Less

Submitted 12 May, 2026; originally announced May 2026.

VoxAfford: Multi-Scale Voxel-Token Fusion for Open-Vocabulary 3D Affordance Detection

Authors: Haowen Sun, Shaolong Zhang, Mingyang Li, Chengzhong Ma, Xinzhe Chen, Qiongjie Cui, Xingyu Chen, Zeyang Liu, Xuguang Lan

Abstract: Open-vocabulary 3D affordance detection requires localizing interaction regions on point clouds given novel affordance descriptions. Recent methods extend multimodal large language models (MLLMs) with special output tokens that are decoded into segmentation masks. However, these tokens are produced through autoregressive generation, which models sequential dependencies rather than spatial neighbor… ▽ More Open-vocabulary 3D affordance detection requires localizing interaction regions on point clouds given novel affordance descriptions. Recent methods extend multimodal large language models (MLLMs) with special output tokens that are decoded into segmentation masks. However, these tokens are produced through autoregressive generation, which models sequential dependencies rather than spatial neighborhood relations, leaving them semantically rich but spatially impoverished for 3D localization. We propose Voxel-enhanced Affordance detection (VoxAfford), which bypasses this bottleneck by injecting multi-scale geometric features from a frozen pre-trained 3D VQVAE encoder into the output tokens after generation. Each output token uses its affordance semantics as a query to retrieve relevant geometric patterns from its paired voxel scale via cross-attention, with a learned compatibility gate controlling the injection strength. The enhanced tokens are then aggregated into a spatially-aware affordance prompt through semantic-conditioned attention and propagated alongside per-point features to generate the final mask. Experiments on open-vocabulary affordance detection tasks show that VoxAfford achieves state-of-the-art performance with approximately an 8% improvement in mIoU, and real robot experiments confirm zero-shot transfer to novel objects. △ Less

Submitted 2 May, 2026; originally announced May 2026.

Adaptive 3D-RoPE: Physics-Aligned Rotary Positional Encoding for Wireless Foundation Models

Authors: Chenyu Zhang, Xinchen Lyu, Chenshan Ren, Shuhan Liu, Qimei Cui

Abstract: Positional encoding plays a pivotal role in determin?ing the extrapolation and generalization performance of wireless foundation models for channel state information (CSI) modeling, latent characterization, and task-specific prediction. However, existing CSI models inherit static or one-dimensional positional priors from natural language and vision architectures, which fundamentally misalign with… ▽ More Positional encoding plays a pivotal role in determin?ing the extrapolation and generalization performance of wireless foundation models for channel state information (CSI) modeling, latent characterization, and task-specific prediction. However, existing CSI models inherit static or one-dimensional positional priors from natural language and vision architectures, which fundamentally misalign with the intrinsic physics of wireless channels by lacking explicit relative decay, collapsing the 3D spatio-temporal-frequency structure, and remaining scenario?rigid. This paper proposes Adaptive 3D-RoPE, a physics-aligned rotary positional encoding that establishes the structural corner?stone for wireless foundation models. The framework integrates a learnable, axis-decoupled 3D frequency bank to explicitly disentangle multi-dimensional phase dependencies, coupled with a lightweight channel-conditioned controller that dynamically modulates the prior via compact global CSI descriptors. This sample-adaptive mechanism transforms positional encoding from a static transformer component into a dynamic, coherence-aware inductive bias to resolve heterogeneous channel physics. Extensive experiments across 100 datasets demonstrate the superiority of the proposed scheme in both scale extrapolation and zero-shot generalization. Compared to the state-of-the-art, our method achieves up to a 10.7 dB reduction in normalized mean square error (NMSE) under 8 times antenna scale extrapolation. Given the same CSI input scales, our method can also improve zero-shot NMSE by 1.07 dB across unseen mobility scenarios and 0.90 dB in low-frequency-to-millimeter-wave tasks. △ Less

Submitted 1 May, 2026; originally announced May 2026.

Comments: 13 pages, 7 figures

Seedance 2.0: Advancing Video Generation for World Complexity

Authors: Team Seedance, De Chen, Liyang Chen, Xin Chen, Ying Chen, Zhuo Chen, Zhuowei Chen, Feng Cheng, Tianheng Cheng, Yufeng Cheng, Mojie Chi, Xuyan Chi, Jian Cong, Qinpeng Cui, Fei Ding, Qide Dong, Yujiao Du, Haojie Duanmu, Junliang Fan, Jiarui Fang, Jing Fang, Zetao Fang, Chengjian Feng, Yu Gao, Diandian Gu , et al. (146 additional authors not shown)

Abstract: Seedance 2.0 is a new native multi-modal audio-video generation model, officially released in China in early February 2026. Compared with its predecessors, Seedance 1.0 and 1.5 Pro, Seedance 2.0 adopts a unified, highly efficient, and large-scale architecture for multi-modal audio-video joint generation. This allows it to support four input modalities: text, image, audio, and video, by integrating… ▽ More Seedance 2.0 is a new native multi-modal audio-video generation model, officially released in China in early February 2026. Compared with its predecessors, Seedance 1.0 and 1.5 Pro, Seedance 2.0 adopts a unified, highly efficient, and large-scale architecture for multi-modal audio-video joint generation. This allows it to support four input modalities: text, image, audio, and video, by integrating one of the most comprehensive suites of multi-modal content reference and editing capabilities available in the industry to date. It delivers substantial, well-rounded improvements across all key sub-dimensions of video and audio generation. In both expert evaluations and public user tests, the model has demonstrated performance on par with the leading levels in the field. Seedance 2.0 supports direct generation of audio-video content with durations ranging from 4 to 15 seconds, with native output resolutions of 480p and 720p. For multi-modal inputs as reference, its current open platform supports up to 3 video clips, 9 images, and 3 audio clips. In addition, we provide Seedance 2.0 Fast version, an accelerated variant of Seedance 2.0 designed to boost generation speed for low-latency scenarios. Seedance 2.0 has delivered significant improvements to its foundational generation capabilities and multi-modal generation performance, bringing an enhanced creative experience for end users. △ Less

Submitted 15 April, 2026; originally announced April 2026.

Comments: Seedance 2.0 Model Card

AffordSim: A Scalable Data Generator and Benchmark for Affordance-Aware Robotic Manipulation

Authors: Mingyang Li, Haofan Xu, Haowen Sun, Xinzhe Chen, Sihua Ren, Liqi Huang, Xinyang Sui, Chenyang Miao, Jiawei Ye, Qiongjie Cui, Zeyang Liu, Xingyu Chen, Xuguang Lan

Abstract: Many everyday robot manipulation skills are affordance-dependent, with success determined by whether the robot contacts the functional object region required by the subsequent action. Current simulation data generators obtain contacts from generic grasp estimators or per-object manual contact annotations, but generic estimators rank stable grasps without task semantics and often select contacts th… ▽ More Many everyday robot manipulation skills are affordance-dependent, with success determined by whether the robot contacts the functional object region required by the subsequent action. Current simulation data generators obtain contacts from generic grasp estimators or per-object manual contact annotations, but generic estimators rank stable grasps without task semantics and often select contacts that are misaligned with the downstream action, while manual contact annotations must be rewritten for each new object and task. To solve these challenges, we introduce AffordSim, a scalable data generator and benchmark that integrates open-vocabulary 3D affordance prediction into simulation-based trajectory generation. Given a natural-language task description, AffordSim synthesizes a task-relevant scene, emits affordance queries, grounds them on object surfaces, samples region-conditioned grasps, and selects executable candidates with motion planning. It further randomizes object pose, texture, lighting, image noise, and cross-viewpoint backgrounds for sim-to-real transfer. We instantiate AffordSim as a 50-task benchmark across diverse manipulation skills, five robot embodiments, and 500+ rigid and articulated objects. AffordSim achieves 93% of the trajectory collection success rate of manual contact annotations on affordance-critical tasks and 89% on hard composite tasks. Vision-language-action policies trained on AffordSim data transfer zero-shot to a real Franka FR3, reaching 24% average success. △ Less

Submitted 11 May, 2026; v1 submitted 13 April, 2026; originally announced April 2026.

Wireless communication empowers online scheduling of partially-observable transportation multi-robot systems in a smart factory

Authors: Yaxin Liao, Qimei Cui, Kwang-Cheng Chen, Xiong Li, Jinlian Chen, Xiyu Zhao, Xiaofeng Tao, Ping Zhang

Abstract: Achieving agile and reconfigurable production flows in smart factories depends on online multi-robot task assignment (MRTA), which requires online collision-free and congestion-free route scheduling of transportation multi-robot systems (T-MRS), e.g., collaborative automatic guided vehicles (AGVs). Due to the real-time operational requirements and dynamic interactions between T-MRS and production… ▽ More Achieving agile and reconfigurable production flows in smart factories depends on online multi-robot task assignment (MRTA), which requires online collision-free and congestion-free route scheduling of transportation multi-robot systems (T-MRS), e.g., collaborative automatic guided vehicles (AGVs). Due to the real-time operational requirements and dynamic interactions between T-MRS and production MRS, online scheduling under partial observability in dynamic factory environments remains a significant and under-explored challenge. This paper proposes a novel communication-enabled online scheduling framework that explicitly couples wireless machine-to-machine (M2M) networking with route scheduling, enabling AGVs to exchange intention information, e.g., planned routes, to overcome partial observations and assist complex computation of online scheduling. Specifically, we determine intelligent AGVs' intention and sensor data as new M2M traffic and tailor the retransmission-free multi-link transmission networking to meet real-time operation demands. This scheduling-oriented networking is then integrated with a simulated annealing-based MRTA scheme and a congestion-aware A*-based route scheduling method. The integrated communication and scheduling scheme allows AGVs to dynamically adjust collision-free and congestion-free routes with reduced computational overhead. Numerical experiments shows the impacts from wireless communication on the performance of T-MRS and suggest that the proposed integrated scheme significantly enhances scheduling efficiency compared to other baselines, even under high AGV load conditions and limited channel resources. Moreover, the results reveal that the scheduling-oriented wireless M2M communication design fundamentally differs from human-to-human communications, implying new technological opportunities in a wireless networked smart factory. △ Less

Submitted 25 March, 2026; originally announced March 2026.

Profound impacts of interlayer interactions in bilayer altermagnetic V2S2O

Authors: Siqi Xu, Qilong Cui, Shaowen Xu, Xianbo Chenwei, Jiahao Zhang, Ruixue Li, Yuan Li, Gaofeng Xu, Fanhao Jia

Abstract: Two-dimensional altermagnets exhibit exceptional potential for low-power spintronics via nonrelativistic spin splitting and zero net magnetization. Here, we systematically investigate the influence of interlayer interactions on the electronic, magnetic and quantum transport properties of bilayer vanadium oxysulfide (V2S2O), a prototypical layered altermagnet, using DFT and NEGF calculations. Our r… ▽ More Two-dimensional altermagnets exhibit exceptional potential for low-power spintronics via nonrelativistic spin splitting and zero net magnetization. Here, we systematically investigate the influence of interlayer interactions on the electronic, magnetic and quantum transport properties of bilayer vanadium oxysulfide (V2S2O), a prototypical layered altermagnet, using DFT and NEGF calculations. Our results reveal that interlayer interactions predominantly modulate the p-orbital derived top valence bands, inducing a profound competitive valence band maximum position between Gamma-point pz and X/Y-point pxy orbitals, with an energy difference as small as 9 meV. Furthermore, interlayer interactions suppress the piezomagnetic effect and impose additional requirements on the type of strain for the bilayer system, compared to its monolayer counterpart. Out-of-plane external electric fields effectively weaken interlayer coupling by enlarging the energy difference of Gamma/X-Y top valence bands to 170 meV. Quantum transport simulations on a bilayer Au/V2S2O/Au two-probe device demonstrate the presence of pronounced spin current. Interlayer interactions reduce the transmission spin polarization from nearly 100% (monolayer) to 60% (bilayer) for energies above the Fermi level. Notably, gate-voltage modulation exhibits significant asymmetry in controlling charge-to-spin current conversion efficiency, originating from the out-of-plane symmetry breaking induced by the electrode geometry. Specifically, a positive gate voltage markedly enhances the contribution of the bottom layer to the overall spin polarization, while a negative gate voltage induces a marginal reduction of transmission spin polarization, attributed to the inherently weak polarization contribution of the bottom layer. These findings provide essential insights for the design and optimization of multilayer altermagnetic spintronics. △ Less

Submitted 24 March, 2026; originally announced March 2026.

Static Scene Reconstruction from Dynamic Egocentric Videos

Authors: Qifei Cui, Patrick Chen

Abstract: Egocentric videos present unique challenges for 3D reconstruction due to rapid camera motion and frequent dynamic interactions. State-of-the-art static reconstruction systems, such as MapAnything, often degrade in these settings, suffering from catastrophic trajectory drift and "ghost" geometry caused by moving hands. We bridge this gap by proposing a robust pipeline that adapts static reconstruct… ▽ More Egocentric videos present unique challenges for 3D reconstruction due to rapid camera motion and frequent dynamic interactions. State-of-the-art static reconstruction systems, such as MapAnything, often degrade in these settings, suffering from catastrophic trajectory drift and "ghost" geometry caused by moving hands. We bridge this gap by proposing a robust pipeline that adapts static reconstruction backbones to long-form egocentric video. Our approach introduces a mask-aware reconstruction mechanism that explicitly suppresses dynamic foreground in the attention layers, preventing hand artifacts from contaminating the static map. Furthermore, we employ a chunked reconstruction strategy with pose-graph stitching to ensure global consistency and eliminate long-term drift. Experiments on HD-EPIC and indoor drone datasets demonstrate that our pipeline significantly improves absolute trajectory error and yields visually clean static geometry compared to naive baselines, effectively extending the capability of foundation models to dynamic first-person scenes. △ Less

Submitted 23 March, 2026; originally announced March 2026.

Waveforms and Fluxes of Generic Extreme-Mass-Ratio Inspirals with a Spinning Secondary

Authors: Qiuxin Cui, Wen-Biao Han

Abstract: Extreme mass-ratio inspirals (EMRIs), comprising a stellar-mass compact object (CO) orbiting a supermassive black hole (BH), are key targets for future space-based gravitational-wave (GW) observatories. Incorporating the spin of the secondary body into waveform models not only enhances measurement precision but also offers insight into the spin distribution of stellar-mass COs. In this work, we co… ▽ More Extreme mass-ratio inspirals (EMRIs), comprising a stellar-mass compact object (CO) orbiting a supermassive black hole (BH), are key targets for future space-based gravitational-wave (GW) observatories. Incorporating the spin of the secondary body into waveform models not only enhances measurement precision but also offers insight into the spin distribution of stellar-mass COs. In this work, we construct the flux and waveform for an EMRI with a spinning secondary in a Kerr background under the linear-spin approximation. Using the radiative prescription (half-retarded minus half-advanced field), we calculate orbit-averaged fluxes for the fundamental constants of motion, including the energy, angular momentum, and the Carter-like constant. This framework provides a tractable route to generating waveforms that incorporate the secondary spin. △ Less

Submitted 14 July, 2026; v1 submitted 18 March, 2026; originally announced March 2026.

Star formation in the circumgalactic high-velocity cloud Complex H

Authors: Zhihong He, Wenkang Pang, Kun Wang, Yangping Luo, Qian Cui

Abstract: The accretion of metal-poor gas sustains galactic star formation. In the Milky Way, this process is fueled by high-velocity clouds (HVCs), yet their fundamental properties have remained elusive in the absence of stellar tracers. Here we report a binary open cluster within HVC Complex H. With an age of 11.2 +- 0.6 Myr and a subsolar metallicity of 0.05(+0.05-0.02) Zsun, the clusters provide a direc… ▽ More The accretion of metal-poor gas sustains galactic star formation. In the Milky Way, this process is fueled by high-velocity clouds (HVCs), yet their fundamental properties have remained elusive in the absence of stellar tracers. Here we report a binary open cluster within HVC Complex H. With an age of 11.2 +- 0.6 Myr and a subsolar metallicity of 0.05(+0.05-0.02) Zsun, the clusters provide a direct stellar distance anchor to the cloud at 13.8 +- 0.6 kpc. Their proper motions indicate Complex H is on a prograde, south-to-north orbit through the outer Galactic disk. The resulting interaction produces a 'slow-fast-slow' velocity gradient, with the cloud's outer layers decelerating as they merge into the disk. Orbit integration suggests the clusters formed from an internal cloud-cloud collision. This triggering mechanism implies other HVCs could similarly produce high-velocity stars. The scarcity of previous stellar detections in HVCs is explained by the rapid escape of young stars (< 20 Myr), while CO non-detections may stem from weak emission due to low metallicity and gas dispersal. This work reveals that the circumgalactic medium can sustain star formation, offering a tangible laboratory to probe the physical conditions of accreting gas before it merges with the Galactic disk. △ Less

Submitted 11 March, 2026; originally announced March 2026.

Comments: 31 pages, 10 figures, published on Nature Astronomy (https://www.nature.com/articles/s41550-026-02814-9)

Commissioning and Full Realization of the PLASEN System at BRIF

Authors: W. C. Mei, H. R. Hu, Y. F. Guo, Z. Yan, X. F. Yang, S. J. Chen, D. Y. Chen, Y. P. Lin, Y. S. Liu, C. Zhang, Y. P. Jing, T. X. Gao, X. Shen, Y. Y. Jia, Y. T. Lin, H. X. Zhang, S. W. Bai, B. Tang, X. Ma, G. F. Song, S. Ye, M. Y. Lu, J. Y. Dong, B. K. Dong, J. H. Lv , et al. (15 additional authors not shown)

Abstract: A PLASEN (Precision LAser Spectroscopy for Exotic Nuclei) system, consisting of a compact radio-frequency quadrupole cooler-buncher (RFQ-cb) and a collinear resonance ionization spectroscopy setup, has now been fully commissioned with radioactive ion beams at the Beijing Radioactive Ion-beam Facility (BRIF). Using both stable and radioactive Rb ion beams from BRIF, we demonstrated that the large b… ▽ More A PLASEN (Precision LAser Spectroscopy for Exotic Nuclei) system, consisting of a compact radio-frequency quadrupole cooler-buncher (RFQ-cb) and a collinear resonance ionization spectroscopy setup, has now been fully commissioned with radioactive ion beams at the Beijing Radioactive Ion-beam Facility (BRIF). Using both stable and radioactive Rb ion beams from BRIF, we demonstrated that the large beam energy spread observed at BRIF has been successfully handled by employing the RFQ-cb, enabling the delivery of high-quality bunched radioactive ion beams for collinear resonance ionization spectroscopy experiments. Under these conditions, we performed laser spectroscopy of exotic nuclei, achieving high resolution (about 100 MHz spectral linewidth) and high sensitivity (up to 1:200 efficiency). This fully operational PLASEN system will serve as a state-of-the-art experimental platform at BRIF for research in multiple fields such as nuclear, atomic and molecular physics. △ Less

Submitted 4 March, 2026; originally announced March 2026.

Journal ref: Phys. Rev. Research 8, 023286 (2026)

GRIP: Geometric Refinement and Adaptive Information Potential for Data Efficiency

Authors: Changhao Wang, Jiaolong Yang, Xinhao Yao, Yunfei Yu, Peng Jiao, Lu Yu, Junpeng Fang, Riccardo Cantoro, Qing Cui, Jun Zhou

Abstract: The performance of Large Language Models (LLMs) is increasingly governed by data efficiency rather than raw scaling volume. However, existing selection methods often decouple global distribution balancing from local instance selection, compromising the hierarchical integrity of the training set. We introduce \textbf{GRIP} (Geometric Refinement and Adaptive Information Potential), a framework that… ▽ More The performance of Large Language Models (LLMs) is increasingly governed by data efficiency rather than raw scaling volume. However, existing selection methods often decouple global distribution balancing from local instance selection, compromising the hierarchical integrity of the training set. We introduce \textbf{GRIP} (Geometric Refinement and Adaptive Information Potential), a framework that unifies these dimensions by modeling the corpus as an information-dense geometric space. GRIP employs a \textbf{Rapid Adaptation Probe (RAP)} to quantify the information potential of semantic clusters, dynamically re-allocating the sampling budget to regions with the highest representation deficits. Subsequently, we perform Intra-Cluster Selection using a \textbf{length-rectified geometric prior} to counteract embedding density artifacts and preserve long-tail logical sequences. Extensive evaluations on Mixture-of-Experts (MoE) models up to 300B tokens demonstrate that GRIP consistently outperforms state-of-the-art baselines, \textbf{surpassing the performance of models trained on $3\times$ larger uncurated datasets}. Our work establishes a robust geometric foundation for adaptive data curation in large-scale pre-training. △ Less

Submitted 4 February, 2026; originally announced March 2026.

Ski Rental with Distributional Predictions of Unknown Quality

Authors: Qiming Cui, Michael Dinitz

Abstract: We revisit the central online problem of ski rental in the "algorithms with predictions" framework from the point of view of distributional predictions. Ski rental was one of the first problems to be studied with predictions, where a natural prediction is simply the number of ski days. But it is both more natural and potentially more powerful to think of a prediction as a distribution p-hat over t… ▽ More We revisit the central online problem of ski rental in the "algorithms with predictions" framework from the point of view of distributional predictions. Ski rental was one of the first problems to be studied with predictions, where a natural prediction is simply the number of ski days. But it is both more natural and potentially more powerful to think of a prediction as a distribution p-hat over the ski days. If the true number of ski days is drawn from some true (but unknown) distribution p, then we show as our main result that there is an algorithm with expected cost at most OPT + O(min(max({eta}, 1) * sqrt(b), b log b)), where OPT is the expected cost of the optimal policy for the true distribution p, b is the cost of buying, and {eta} is the Earth Mover's (Wasserstein-1) distance between p and p-hat. Note that when {eta} < o(sqrt(b)) this gives additive loss less than b (the trivial bound), and when {eta} is arbitrarily large (corresponding to an extremely inaccurate prediction) we still do not pay more than O(b log b) additive loss. An implication of these bounds is that our algorithm has consistency O(sqrt(b)) (additive loss when the prediction error is 0) and robustness O(b log b) (additive loss when the prediction error is arbitrarily large). Moreover, we do not need to assume that we know (or have any bound on) the prediction error {eta}, in contrast with previous work in robust optimization which assumes that we know this error. We complement this upper bound with a variety of lower bounds showing that it is essentially tight: not only can the consistency/robustness tradeoff not be improved, but our particular loss function cannot be meaningfully improved. △ Less

Submitted 24 February, 2026; originally announced February 2026.

On Representation Redundancy in Large-Scale Instruction Tuning Data Selection

Authors: Youwei Shu, Shaomian Zheng, Dingnan Jin, Wenjie Qu, Ziyao Guo, Qing Cui, Jun Zhou, Jiaheng Zhang

Abstract: Data quality is a crucial factor in large language models training. While prior work has shown that models trained on smaller, high-quality datasets can outperform those trained on much larger but noisy or low-quality corpora, systematic methods for industrial-scale data selection in instruction tuning remain underexplored. In this work, we study instruction-tuning data selection through the lens… ▽ More Data quality is a crucial factor in large language models training. While prior work has shown that models trained on smaller, high-quality datasets can outperform those trained on much larger but noisy or low-quality corpora, systematic methods for industrial-scale data selection in instruction tuning remain underexplored. In this work, we study instruction-tuning data selection through the lens of semantic representation similarity and identify a key limitation of state-of-the-art LLM encoders: they produce highly redundant semantic embeddings. To mitigate this redundancy, we propose Compressed Representation Data Selection (CRDS), a novel framework with two variants. CRDS-R applies Rademacher random projection followed by concatenation of transformer hidden-layer representations, while CRDS-W employs whitening-based dimensionality reduction to improve representational quality. Experimental results demonstrate that both variants substantially enhance data quality and consistently outperform state-of-the-art representation-based selection methods. Notably, CRDS-W achieves strong performance using only 3.5% of the data, surpassing the full-data baseline by an average of 0.71% across four datasets. Our code is available at https://github.com/tdano1/CRDS. △ Less

Submitted 14 February, 2026; originally announced February 2026.

Mergers Drive Structural Complexity but Not Starbursts in Lyman-$α$ Emitters at $3 < z < 4$: A JWST Spatially Resolved View

Authors: Qi Song, F. S. Liu, Jian Ren, Tianfu Gao, Pinsong Zhao, Qifan Cui, Yubin Li, Hao Mo, Guanghuan Wang

Abstract: Recent observations with the James Webb Space Telescope (JWST) reveal that the merger fraction among Ly$α$ emitters (LAEs) at redshifts $z > 3$ is significantly higher than previously estimated. In this study, we focus on three high signal-to-noise merging LAE systems at $3 < z < 4$, selected from the VLT/MUSE-Deep survey in the GOODS-S field. We combine new \textit{JWST}/NIRCam broadband and medi… ▽ More Recent observations with the James Webb Space Telescope (JWST) reveal that the merger fraction among Ly$α$ emitters (LAEs) at redshifts $z > 3$ is significantly higher than previously estimated. In this study, we focus on three high signal-to-noise merging LAE systems at $3 < z < 4$, selected from the VLT/MUSE-Deep survey in the GOODS-S field. We combine new \textit{JWST}/NIRCam broadband and medium-band imaging with archival \textit{HST}/ACS data to perform spatially resolved spectral energy distribution (SED) fitting using the \textsc{Bagpipes} software package. Our analysis reveals that two of the systems are minor mergers, while the third is a major merger. The close agreement between spatially resolved and integrated stellar mass estimates indicates that recent star formation does not significantly outshine the light from older stellar populations in these systems. Moreover, both the individual components and the systems as a whole lie on the star-forming main sequence, further supporting the conclusion that these mergers have not yet triggered substantial starburst activity. Furthermore, we detect prominent color gradients and disturbed dust distributions in these merging systems, indicating that the mergers have already induced significant internal structural perturbations. These morphological and dust-related changes may facilitate the escape of Ly$α$ photons -- potentially through mechanisms such as gas redistribution or a reduced covering fraction of neutral hydrogen -- thereby playing a key role in shaping the observed properties of LAEs. △ Less

Submitted 5 February, 2026; originally announced February 2026.

Comments: Accepted for publication in RAA

UniGeM: Unifying Data Mixing and Selection via Geometric Exploration and Mining

Authors: Changhao Wang, Yunfei Yu, Xinhao Yao, Jiaolong Yang, Riccardo Cantoro, Chaobo Li, Qing Cui, Jun Zhou

Abstract: The scaling of Large Language Models (LLMs) is increasingly limited by data quality. Most methods handle data mixing and sample selection separately, which can break the structure in code corpora. We introduce \textbf{UniGeM}, a framework that unifies mixing and selection by treating data curation as a \textit{manifold approximation} problem without training proxy models or relying on external ref… ▽ More The scaling of Large Language Models (LLMs) is increasingly limited by data quality. Most methods handle data mixing and sample selection separately, which can break the structure in code corpora. We introduce \textbf{UniGeM}, a framework that unifies mixing and selection by treating data curation as a \textit{manifold approximation} problem without training proxy models or relying on external reference datasets. UniGeM operates hierarchically: \textbf{Macro-Exploration} learns mixing weights with stability-based clustering; \textbf{Micro-Mining} filters high-quality instances by their geometric distribution to ensure logical consistency. Validated by training 8B and 16B MoE models on 100B tokens, UniGeM achieves \textbf{2.0$\times$ data efficiency} over a random baseline and further improves overall performance compared to SOTA methods in reasoning-heavy evaluations and multilingual generalization. △ Less

Submitted 3 February, 2026; originally announced February 2026.

Coverage Performance Analysis of FAS-enhanced LoRa Wide Area Networks under both Co-SF and Inter-SF Interference

Authors: Gaoze Mu, Yanzhao Hou, Mingjie Chen, Yuanyu Hu, Yongan Zheng, Qimei Cui, Xiaofeng Tao

Abstract: This paper presents an analytical framework for evaluating the coverage performance of the fluid antenna system (FAS)-enhanced LoRa wide-area networks (LoRaWANs). We investigate the effects of large-scale pathloss in LoRaWAN, small-scale fading characterized by FAS, and dense interference (i.e., packet collisions under the ALOHA protocol) arising from randomly deployed end devices (EDs). Both co-s… ▽ More This paper presents an analytical framework for evaluating the coverage performance of the fluid antenna system (FAS)-enhanced LoRa wide-area networks (LoRaWANs). We investigate the effects of large-scale pathloss in LoRaWAN, small-scale fading characterized by FAS, and dense interference (i.e., packet collisions under the ALOHA protocol) arising from randomly deployed end devices (EDs). Both co-spreading factor (co-SF) interference (with the same SF) and inter-SF interference (with different SFs) are introduced into the network, and their differences in physical characteristics are also considered in the analysis. Additionally, simple yet accurate statistical approximations of the FAS channel envelope and power are derived using the extreme-value theorem. Based on the approximated channel expression, the theoretical coverage probability of the proposed FAS-enhanced LoRaWAN is derived. Numerical results validate our analytical approximations by exhibiting close agreement with the exact correlation model. Notably, it is revealed that a FAS with a normalized aperture of 1 times 1 can greatly enhance network performance, in terms of both ED numbers and coverage range. △ Less

Submitted 19 April, 2026; v1 submitted 27 January, 2026; originally announced January 2026.

Comments: 6 pages, 3 figures

HeterCSI: Channel-Adaptive Heterogeneous CSI Pretraining Framework for Generalized Wireless Foundation Models

Authors: Chenyu Zhang, Xinchen Lyu, Chenshan Ren, Shuhan Liu, Qimei Cui, Xiaofeng Tao

Abstract: Wireless foundation models promise transformative capabilities for channel state information (CSI) processing across diverse 6G network applications, yet face fundamental challenges due to the inherent dual heterogeneity of CSI across both scale and scenario dimensions. However, current pretraining approaches either constrain inputs to fixed dimensions or isolate training by scale, limiting the ge… ▽ More Wireless foundation models promise transformative capabilities for channel state information (CSI) processing across diverse 6G network applications, yet face fundamental challenges due to the inherent dual heterogeneity of CSI across both scale and scenario dimensions. However, current pretraining approaches either constrain inputs to fixed dimensions or isolate training by scale, limiting the generalization and scalability of wireless foundation models. In this paper, we propose HeterCSI, a channel-adaptive pretraining framework that reconciles training efficiency with robust cross-scenario generalization via a new understanding of gradient dynamics in heterogeneous CSI pretraining. Our key insight reveals that CSI scale heterogeneity primarily causes destructive gradient interference, while scenario diversity actually promotes constructive gradient alignment when properly managed. Specifically, we formulate heterogeneous CSI batch construction as a partitioning optimization problem that minimizes zero-padding overhead while preserving scenario diversity. To solve this, we develop a scale-aware adaptive batching strategy that aligns CSI samples of similar scales, and design a double-masking mechanism to isolate valid signals from padding artifacts. Extensive experiments on 12 datasets demonstrate that HeterCSI establishes a generalized foundation model without scenario-specific finetuning, achieving superior average performance over full-shot baselines. Compared to the state-of-the-art zero-shot benchmark WiFo, it reduces NMSE by 7.19 dB, 4.08 dB, and 5.27 dB for CSI reconstruction, time-domain, and frequency-domain prediction, respectively. The proposed HeterCSI framework also reduces training latency by 53% compared to existing approaches while improving generalization performance by 1.53 dB on average. △ Less

Submitted 26 January, 2026; originally announced January 2026.

Comments: 13 pages, 8 figures

Multicultural Spyfall: Assessing LLMs through Dynamic Multilingual Social Deduction Game

Authors: Haryo Akbarianto Wibowo, Alaa Elsetohy, Qinrong Cui, Alham Fikri Aji

Abstract: The rapid advancement of Large Language Models (LLMs) has necessitated more robust evaluation methods that go beyond static benchmarks, which are increasingly prone to data saturation and leakage. In this paper, we propose a dynamic benchmarking framework for evaluating multilingual and multicultural capabilities through the social deduction game Spyfall. In our setup, models must engage in strate… ▽ More The rapid advancement of Large Language Models (LLMs) has necessitated more robust evaluation methods that go beyond static benchmarks, which are increasingly prone to data saturation and leakage. In this paper, we propose a dynamic benchmarking framework for evaluating multilingual and multicultural capabilities through the social deduction game Spyfall. In our setup, models must engage in strategic dialogue to either identify a secret agent or avoid detection, utilizing culturally relevant locations or local foods. Our results show that our game-based rankings align closely with the Chatbot Arena. However, we find a significant performance gap in non-English contexts: models are generally less proficient when handling locally specific entities and often struggle with rule-following or strategic integrity in non-English languages. We demonstrate that this game-based approach provides a scalable, leakage-resistant, and culturally nuanced alternative to traditional NLP benchmarks. The game history can be accessed here https://huggingface.co/datasets/haryoaw/cultural-spyfall. △ Less

Submitted 13 January, 2026; originally announced January 2026.

MSC Class: 68T50

Reflection-Driven Self-Optimization 6G Agentic AI RAN via Simulation-in-the-Loop Workflows

Authors: Yunhao Hu, Xinchen Lyu, Chenshan Ren, Keda Chen, Qimei Cui, Xiaofeng Tao

Abstract: The escalating complexity of sixth-generation (6G) networks demands unprecedented levels of autonomy beyond the capabilities of traditional optimization-based and current AI-based resource management approaches. While agentic AI has emerged as a promising paradigm for autonomous RAN, current frameworks provide sophisticated reasoning capabilities but lack mechanisms for empirical validation and se… ▽ More The escalating complexity of sixth-generation (6G) networks demands unprecedented levels of autonomy beyond the capabilities of traditional optimization-based and current AI-based resource management approaches. While agentic AI has emerged as a promising paradigm for autonomous RAN, current frameworks provide sophisticated reasoning capabilities but lack mechanisms for empirical validation and self-improvement. This article identifies simulation-in-the-loop validation as a critical enabler for truly autonomous networks, where AI agents can empirically verify decisions and learn from outcomes. We present the first reflection-driven self-optimization framework that integrates agentic AI with high-fidelity network simulation in a closed-loop architecture. Our system orchestrates four specialized agents, including scenario, solver, simulation, and reflector agents, working in concert to transform agentic AI into a self-correcting system capable of escaping local optima, recognizing implicit user intent, and adapting to dynamic network conditions. Extensive experiments validate significant performance improvements over non-agentic approaches: 17.1\% higher throughput in interference optimization, 67\% improved user QoS satisfaction through intent recognition, and 25\% reduced resource utilization during low-traffic periods while maintaining service quality. △ Less

Submitted 21 April, 2026; v1 submitted 8 December, 2025; originally announced December 2025.

Seedance 1.5 pro: A Native Audio-Visual Joint Generation Foundation Model

Authors: Team Seedance, Heyi Chen, Siyan Chen, Xin Chen, Yanfei Chen, Ying Chen, Zhuo Chen, Feng Cheng, Tianheng Cheng, Xinqi Cheng, Xuyan Chi, Jian Cong, Jing Cui, Qinpeng Cui, Qide Dong, Junliang Fan, Jing Fang, Zetao Fang, Chengjian Feng, Han Feng, Mingyuan Gao, Yu Gao, Dong Guo, Qiushan Guo, Boyang Hao , et al. (172 additional authors not shown)

Abstract: Recent strides in video generation have paved the way for unified audio-visual generation. In this work, we present Seedance 1.5 pro, a foundational model engineered specifically for native, joint audio-video generation. Leveraging a dual-branch Diffusion Transformer architecture, the model integrates a cross-modal joint module with a specialized multi-stage data pipeline, achieving exceptional au… ▽ More Recent strides in video generation have paved the way for unified audio-visual generation. In this work, we present Seedance 1.5 pro, a foundational model engineered specifically for native, joint audio-video generation. Leveraging a dual-branch Diffusion Transformer architecture, the model integrates a cross-modal joint module with a specialized multi-stage data pipeline, achieving exceptional audio-visual synchronization and superior generation quality. To ensure practical utility, we implement meticulous post-training optimizations, including Supervised Fine-Tuning (SFT) on high-quality datasets and Reinforcement Learning from Human Feedback (RLHF) with multi-dimensional reward models. Furthermore, we introduce an acceleration framework that boosts inference speed by over 10X. Seedance 1.5 pro distinguishes itself through precise multilingual and dialect lip-syncing, dynamic cinematic camera control, and enhanced narrative coherence, positioning it as a robust engine for professional-grade content creation. Seedance 1.5 pro is now accessible on Volcano Engine at https://console.volcengine.com/ark/region:ark+cn-beijing/experience/vision?type=GenVideo. △ Less

Submitted 23 December, 2025; v1 submitted 15 December, 2025; originally announced December 2025.

Comments: Seedance 1.5 pro Technical Report

Advancing LLM-Based Security Automation with Customized Group Relative Policy Optimization for Zero-Touch Networks

Authors: Xinye Cao, Yihan Lin, Guoshun Nan, Qinchuan Zhou, Yuhang Luo, Yurui Gao, Zeliang Zhang, Haolang Lu, Qimei Cui, Yanzhao Hou, Xiaofeng Tao, Tony Q. S. Quek

Abstract: Zero-Touch Networks (ZTNs) represent a transformative paradigm toward fully automated and intelligent network management, providing the scalability and adaptability required for the complexity of sixth-generation (6G) networks. However, the distributed architecture, high openness, and deep heterogeneity of 6G networks expand the attack surface and pose unprecedented security challenges. To address… ▽ More Zero-Touch Networks (ZTNs) represent a transformative paradigm toward fully automated and intelligent network management, providing the scalability and adaptability required for the complexity of sixth-generation (6G) networks. However, the distributed architecture, high openness, and deep heterogeneity of 6G networks expand the attack surface and pose unprecedented security challenges. To address this, security automation aims to enable intelligent security management across dynamic and complex environments, serving as a key capability for securing 6G ZTNs. Despite its promise, implementing security automation in 6G ZTNs presents two primary challenges: 1) automating the lifecycle from security strategy generation to validation and update under real-world, parallel, and adversarial conditions, and 2) adapting security strategies to evolving threats and dynamic environments. This motivates us to propose SecLoop and SA-GRPO. SecLoop constitutes the first fully automated framework that integrates large language models (LLMs) across the entire lifecycle of security strategy generation, orchestration, response, and feedback, enabling intelligent and adaptive defenses in dynamic network environments, thus tackling the first challenge. Furthermore, we propose SA-GRPO, a novel security-aware group relative policy optimization algorithm that iteratively refines security strategies by contrasting group feedback collected from parallel SecLoop executions, thereby addressing the second challenge. Extensive real-world experiments on five benchmarks, including 11 MITRE ATT&CK processes and over 20 types of attacks, demonstrate the superiority of the proposed SecLoop and SA-GRPO. We will release our platform to the community, facilitating the advancement of security automation towards next generation communications. △ Less

Submitted 10 December, 2025; originally announced December 2025.

Comments: Accepted by IEEE JSAC. This work has been submitted to the IEEE for possible publication

Sensing and Understanding the World over Air: A Large Multimodal Model for Mobile Networks

Authors: Zhuoran Duan, Yuhao Wei, Guoshun Nan, Zijun Wang, Yan Yan, Lihua Xiong, Yuhan Ran, Ji Zhang, Jian Li, Qimei Cui, Xiaofeng Tao, Tony Q. S. Quek

Abstract: Large models (LMs), such as ChatGPT, have made a significant impact across diverse domains and hold great potential to facilitate the evolution of network intelligence. Wireless-native multi-modal large models (WMLMs) can sense and understand the physical world through multi-modal data, serving as a key enabler that integrates communication, sensing, and intelligence, and thus they can boost vario… ▽ More Large models (LMs), such as ChatGPT, have made a significant impact across diverse domains and hold great potential to facilitate the evolution of network intelligence. Wireless-native multi-modal large models (WMLMs) can sense and understand the physical world through multi-modal data, serving as a key enabler that integrates communication, sensing, and intelligence, and thus they can boost various smart services to billions of users. However, research on WMLMs remains in its infancy, and the construction of domain-specific multi-modal large models for wireless networks is still underexplored. In this paper, we outlines the key characteristics of WMLMs and summarizes existing methods, on the basis of which a wireless-native multimodal training paradigm is proposed. Specifically, we constructed a GPT-style WMLM model and trained it on a real-world large-scale dataset, leveraging wireless signals as an anchor modality for contrastive learning. Our approach demonstrates outstanding performance compared with existing small-scale models and large multi-modal models, validating the feasibility of using wireless signals as a universal modality and highlighting WMLM's potential to emerge as a new paradigm for future wireless networks. △ Less

Submitted 17 November, 2025; originally announced November 2025.

An approach of deep reinforcement learning for maximizing the net present value of stochastic projects

Authors: Wei Xu, Fan Yang, Qinyuan Cui, Zhi Chen

Abstract: This paper investigates a project with stochastic activity durations and cash flows under discrete scenarios, where activities must satisfy precedence constraints generating cash inflows and outflows. The objective is to maximize expected net present value (NPV) by accelerating inflows and deferring outflows. We formulate the problem as a discrete-time Markov Decision Process (MDP) and propose a D… ▽ More This paper investigates a project with stochastic activity durations and cash flows under discrete scenarios, where activities must satisfy precedence constraints generating cash inflows and outflows. The objective is to maximize expected net present value (NPV) by accelerating inflows and deferring outflows. We formulate the problem as a discrete-time Markov Decision Process (MDP) and propose a Double Deep Q-Network (DDQN) approach. Comparative experiments demonstrate that DDQN outperforms traditional rigid and dynamic strategies, particularly in large-scale or highly uncertain environments, exhibiting superior computational capability, policy reliability, and adaptability. Ablation studies further reveal that the dual-network architecture mitigates overestimation of action values, while the target network substantially improves training convergence and robustness. These results indicate that DDQN not only achieves higher expected NPV in complex project optimization but also provides a reliable framework for stable and effective policy implementation. △ Less

Submitted 16 November, 2025; originally announced November 2025.

Improving Autoformalization Using Direct Dependency Retrieval

Authors: Shaoqi Wang, Lu Yu, Siwei Lou, Feng Yan, Chunjie Yang, Qing Cui, Jun Zhou

Abstract: The convergence of deep learning and formal mathematics has spurred research in formal verification. Statement autoformalization, a crucial first step in this process, aims to translate informal descriptions into machine-verifiable representations but remains a significant challenge. The core difficulty lies in the fact that existing methods often suffer from a lack of contextual awareness, leadin… ▽ More The convergence of deep learning and formal mathematics has spurred research in formal verification. Statement autoformalization, a crucial first step in this process, aims to translate informal descriptions into machine-verifiable representations but remains a significant challenge. The core difficulty lies in the fact that existing methods often suffer from a lack of contextual awareness, leading to hallucination of formal definitions and theorems. Furthermore, current retrieval-augmented approaches exhibit poor precision and recall for formal library dependency retrieval, and lack the scalability to effectively leverage ever-growing public datasets. To bridge this gap, we propose a novel retrieval-augmented framework based on DDR (\textit{Direct Dependency Retrieval}) for statement autoformalization. Our DDR method directly generates candidate library dependencies from natural language mathematical descriptions and subsequently verifies their existence within the formal library via an efficient suffix array check. Leveraging this efficient search mechanism, we constructed a dependency retrieval dataset of over 500,000 samples and fine-tuned a high-precision DDR model. Experimental results demonstrate that our DDR model significantly outperforms SOTA methods in both retrieval precision and recall. Consequently, an autoformalizer equipped with DDR shows consistent performance advantages in both single-attempt accuracy and multi-attempt stability compared to models using traditional selection-based RAG methods. △ Less

Submitted 1 January, 2026; v1 submitted 14 November, 2025; originally announced November 2025.

Every Activation Boosted: Scaling General Reasoner to 1 Trillion Open Language Foundation

Authors: Ling Team, Ang Li, Ben Liu, Binbin Hu, Bing Li, Bingwei Zeng, Borui Ye, Caizhi Tang, Changxin Tian, Chao Huang, Chao Zhang, Chen Qian, Chenchen Ju, Chenchen Li, Chengfu Tang, Chilin Fu, Chunshao Ren, Chunwei Wu, Cong Zhang, Cunyin Peng, Dafeng Xu, Daixin Wang, Dalong Zhang, Dingnan Jin, Dingyuan Zhu , et al. (117 additional authors not shown)

Abstract: We introduce Ling 2.0, a series reasoning-oriented language foundation built upon the principle that every activation boosts reasoning capability. Designed to scale from tens of billions to one trillion parameters under a unified Mixture-of-Experts (MoE) paradigm, Ling 2.0 emphasizes high sparsity, cross-scale consistency, and efficiency guided by empirical scaling laws. The series includes three… ▽ More We introduce Ling 2.0, a series reasoning-oriented language foundation built upon the principle that every activation boosts reasoning capability. Designed to scale from tens of billions to one trillion parameters under a unified Mixture-of-Experts (MoE) paradigm, Ling 2.0 emphasizes high sparsity, cross-scale consistency, and efficiency guided by empirical scaling laws. The series includes three non-thinking (instruct) models - Ling-mini-2.0, Ling-flash-2.0, and Ling-1T - ranging from 16B to 1T total parameters and achieving up to 7-fold active-compute efficiency compared with dense counterparts. Ling 2.0 integrates coordinated innovations across model architecture, pre-training, post-training, and infrastructure: a high-sparsity MoE with MTP for efficient reasoning, reasoning-oriented data and mid-training CoT activation, reinforcement-based fine-tuning (DFT, Evo-CoT), and full-scale FP8 training with fine-grained heterogeneous pipelines. At the trillion scale, Ling-1T establishes a new Pareto frontier of reasoning accuracy versus computational efficiency, demonstrating that sparse activation, when properly aligned with reasoning objectives, enables scalable and efficient intelligence. Collectively, Ling 2.0 provides a coherent, open, and efficient foundation for advancing future reasoning and thinking models, including the Ring series built upon the same base. △ Less

Submitted 6 November, 2025; v1 submitted 24 October, 2025; originally announced October 2025.

Comments: Ling 2.0 Technical Report

See, Think, Act: Online Shopper Behavior Simulation with VLM Agents

Authors: Yimeng Zhang, Jiri Gesi, Ran Xue, Tian Wang, Ziyi Wang, Yuxuan Lu, Sinong Zhan, Huimin Zeng, Qingjun Cui, Yufan Guo, Jing Huang, Mubarak Shah, Dakuo Wang

Abstract: LLMs have recently demonstrated strong potential in simulating online shopper behavior. Prior work has improved action prediction by applying SFT on action traces with LLM-generated rationales, and by leveraging RL to further enhance reasoning capabilities. Despite these advances, current approaches rely on text-based inputs and overlook the essential role of visual perception in shaping human dec… ▽ More LLMs have recently demonstrated strong potential in simulating online shopper behavior. Prior work has improved action prediction by applying SFT on action traces with LLM-generated rationales, and by leveraging RL to further enhance reasoning capabilities. Despite these advances, current approaches rely on text-based inputs and overlook the essential role of visual perception in shaping human decision-making during web GUI interactions. In this paper, we investigate the integration of visual information, specifically webpage screenshots, into behavior simulation via VLMs, leveraging OPeRA dataset. By grounding agent decision-making in both textual and visual modalities, we aim to narrow the gap between synthetic agents and real-world users, thereby enabling more cognitively aligned simulations of online shopping behavior. Specifically, we employ SFT for joint action prediction and rationale generation, conditioning on the full interaction context, which comprises action history, past HTML observations, and the current webpage screenshot. To further enhance reasoning capabilities, we integrate RL with a hierarchical reward structure, scaled by a difficulty-aware factor that prioritizes challenging decision points. Empirically, our studies show that incorporating visual grounding yields substantial gains: the combination of text and image inputs improves exact match accuracy by more than 6% over text-only inputs. These results indicate that multi-modal grounding not only boosts predictive accuracy but also enhances simulation fidelity in visually complex environments, which captures nuances of human attention and decision-making that text-only agents often miss. Finally, we revisit the design space of behavior simulation frameworks, identify key methodological limitations, and propose future research directions toward building efficient and effective human behavior simulators. △ Less

Submitted 22 October, 2025; originally announced October 2025.

FourierCompress: Layer-Aware Spectral Activation Compression for Efficient and Accurate Collaborative LLM Inference

Authors: Jian Ma, Xinchen Lyu, Jun Jiang, Longhao Zou, Chenshan Ren, Qimei Cui, Xiaofeng Tao

Abstract: Collaborative large language model (LLM) inference enables real-time, privacy-preserving AI services on resource-constrained edge devices by partitioning computational workloads between client devices and edge servers. However, this paradigm is severely hindered by communication bottlenecks caused by the transmission of high-dimensional intermediate activations, exacerbated by the autoregressive d… ▽ More Collaborative large language model (LLM) inference enables real-time, privacy-preserving AI services on resource-constrained edge devices by partitioning computational workloads between client devices and edge servers. However, this paradigm is severely hindered by communication bottlenecks caused by the transmission of high-dimensional intermediate activations, exacerbated by the autoregressive decoding structure of LLMs, where bandwidth consumption scales linearly with output length. Existing activation compression methods struggle to simultaneously achieve high compression ratios, low reconstruction error, and computational efficiency. This paper proposes FourierCompress, a novel, layer-aware activation compression framework that exploits the frequency-domain sparsity of LLM activations. We rigorously demonstrate that activations from the first Transformer layer exhibit strong smoothness and energy concentration in the low-frequency domain, making them highly amenable to near-lossless compression via the Fast Fourier Transform (FFT). FourierCompress transforms activations into the frequency domain, retains only a compact block of low-frequency coefficients, and reconstructs the signal at the server using conjugate symmetry, enabling seamless hardware acceleration on DSPs and FPGAs. Extensive experiments on Llama 3 and Qwen2.5 models across 10 commonsense reasoning datasets demonstrate that FourierCompress preserves performance remarkably close to the uncompressed baseline, outperforming Top-k, QR, and SVD. FourierCompress bridges the gap between communication efficiency (an average 7.6x reduction in activation size), near-lossless inference (less than 0.3% average accuracy loss), and significantly faster compression (achieving over 32x reduction in compression time compared to Top-k via hardware acceleration) for edge-device LLM inference. △ Less

Submitted 18 October, 2025; originally announced October 2025.

Some results on minimum saturated graphs

Authors: Chenke Zhang, Qing Cui, Jinze Hu, Erfei Yue, Shengjin Ji

Abstract: Let $G$ be a graph and $\mathcal{F}$ be a family of graphs. We say a graph $G$ is $\mathcal{F}$-saturated if $G$ does not contain any member in $\mathcal{F}$ and for any $e\in E(\overline{G})$, $G+e$ creates a copy of some member in $ \mathcal{F}$. The saturation number of $\mathcal{F}$ is the minimum number of edges of an $\mathcal{F}$-saturated graphs with $n$ vertices, denoted by… ▽ More Let $G$ be a graph and $\mathcal{F}$ be a family of graphs. We say a graph $G$ is $\mathcal{F}$-saturated if $G$ does not contain any member in $\mathcal{F}$ and for any $e\in E(\overline{G})$, $G+e$ creates a copy of some member in $ \mathcal{F}$. The saturation number of $\mathcal{F}$ is the minimum number of edges of an $\mathcal{F}$-saturated graphs with $n$ vertices, denoted by $\sat(n,\mathcal{F})$. If $\mathcal{F}=\{F\}$, then we write it as $\sat(n,F)$ for short. In this paper, we determine the exact value of $\sat(n,\{K_3,P_k\})$, and as its application, we obtain two bounds of $\sat(n,K_3\cup P_k)$ for $k\ge 10$ and sufficiently large $n$. Furthermore, $\sat(n,K_1\lor F)$ is determined, where $F$ is a linear forest without isolated vertices. △ Less

Submitted 12 October, 2025; originally announced October 2025.

Comments: 16 pages,5 figures

Gaia DR3 Open Cluster Cepheids: A Unified Catalog with Calibrated Period-Age and Period-Wesenheit Relations

Authors: Shunhong Deng, Zhihong He, Anbing Ren, Qian Cui, Xiaoyue Zhou, Liming Peng, Chenxin Wang, Ziang Chen, Yangping Luo, Kun Wang

Abstract: Classical Cepheids (CCs) in Galactic open clusters (OCs) provide essential observational constraints for calibrating the period-age relation (PAR) and the period-Wesenheit relation (PWR) of CCs. However, distant and long-period OC Cepheids remain limited, while the confirmed samples still require more precise determinations of their physical properties, such as ages and extinctions. In this work,… ▽ More Classical Cepheids (CCs) in Galactic open clusters (OCs) provide essential observational constraints for calibrating the period-age relation (PAR) and the period-Wesenheit relation (PWR) of CCs. However, distant and long-period OC Cepheids remain limited, while the confirmed samples still require more precise determinations of their physical properties, such as ages and extinctions. In this work, we present a comprehensive census of OC Cepheids based on an extensive sample of distant OCs from Gaia Data Release 3 (DR3). By combining astrometric and photometric membership analyses, we identified 110 CCs associated with 102 OCs, of which 41 CCs across 37 OCs were classified as OC Cepheids, while the remaining cases were considered candidate or rejected associations. Our results are consistent with previous studies, while 4 of the 41 OC Cepheids are newly reported here. Using updated cluster parameters derived from manual isochrone fitting, we primarily refined the PAR to log Age = (-0.595 $\pm$ 0.044) log P + (8.430 $\pm$ 0.042) and recalibrated the PWR to WG = (-3.615 $\pm$ 0.083) log P + (-2.379 $\pm$ 0.096). This study expands the sample of confirmed and candidate OC Cepheids. The newly longest-period confirmed OC Cepheid is BM Per (CWNU 3123) with log P = 1.36, and two newly discovered OC Cepheid candidates have distances exceeding 6 kpc. Moreover, the PAR and PWR are improved by incorporating refined OC ages and updated parallaxes, respectively. △ Less

Submitted 31 October, 2025; v1 submitted 9 October, 2025; originally announced October 2025.

Comments: 29 pages, 6 figures, 3 tables, 6 appendix figures, 3 appendix tables, Accepted for publication in AJ, minor corrections for typo

Learning Function-to-Function Mappings: A Fourier Neural Operator for Next-Generation MIMO Systems

Authors: Jian Xiao, Ji Wang, Qi Sun, Qimei Cui, Xingwang Li, Dusit Niyato, Chih-Lin I

Abstract: Next-generation multiple-input multiple-output (MIMO) systems, characterized by extremely large-scale arrays, holographic surfaces, three-dimensional architectures, and flexible antennas, are poised to deliver unprecedented data rates, spectral efficiency and stability. However, these advancements introduce significant challenges for physical layer signal processing, stemming from complex near-fie… ▽ More Next-generation multiple-input multiple-output (MIMO) systems, characterized by extremely large-scale arrays, holographic surfaces, three-dimensional architectures, and flexible antennas, are poised to deliver unprecedented data rates, spectral efficiency and stability. However, these advancements introduce significant challenges for physical layer signal processing, stemming from complex near-field propagation, continuous aperture modeling, sub-wavelength antenna coupling effects, and dynamic channel conditions. Conventional model-based and deep learning approaches often struggle with the immense computational complexity and model inaccuracies inherent in these new regimes. This article proposes a Fourier neural operator (FNO) as a powerful and promising tool to address these challenges. The FNO learns function-to-function mappings between infinite-dimensional function spaces, making them exceptionally well-suited for modeling complex physical systems governed by partial differential equations based on electromagnetic wave propagation. We first present the fundamental principles of FNO, demonstrating its mesh-free nature and function-to-function ability to efficiently capture global dependencies in the Fourier domain. Furthermore, we explore a range of applications of FNO in physical-layer signal processing for next-generation MIMO systems. Representative case studies on channel modeling and estimation for novel MIMO architectures demonstrate the superior performance of FNO compared to state-of-the-art methods. Finally, we discuss open challenges and outline future research directions, positioning FNO as a promising technology for enabling the enormous potential of next-generation MIMO systems. △ Less

Submitted 6 October, 2025; originally announced October 2025.

The Debate on RLVR Reasoning Capability Boundary: Shrinkage, Expansion, or Both? A Two-Stage Dynamic View

Authors: Xinhao Yao, Lu Yu, Xiaolin Hu, Fengwei Teng, Qing Cui, Jun Zhou, Yong Liu

Abstract: The ongoing debate on whether reinforcement learning with verifiable rewards (RLVR) expands or shrinks the reasoning capabilities of large language models (LLMs) remains unresolved. Some studies contend that RLVR mainly improves sampling efficiency but at the expense of diversity and exploratory capacity, resulting in capability boundary shrinkage. In contrast, others demonstrate that prolonged tr… ▽ More The ongoing debate on whether reinforcement learning with verifiable rewards (RLVR) expands or shrinks the reasoning capabilities of large language models (LLMs) remains unresolved. Some studies contend that RLVR mainly improves sampling efficiency but at the expense of diversity and exploratory capacity, resulting in capability boundary shrinkage. In contrast, others demonstrate that prolonged training can lead to the emergence of novel reasoning strategies, suggesting capability boundary expansion. To reconcile these contradictory findings, we theoretically and empirically show that both perspectives are partially valid-each aligning with a separate phase in an inherent two-stage probability mass dynamic: (1) Exploitation stage: initially, the model primarily samples explored high-reward and low-reward tokens, while rarely selecting the potentially optimal token. Positive advantage estimates increase the probability of high-reward tokens and decrease those of low-reward tokens, yet the optimal token's probability remains largely unchanged during this stage. (2) Exploration stage: as training advances, the growth rate of previously acquired high-reward tokens slows as their probabilities approach saturation. When a potentially optimal token-now receiving positive advantage estimates-is occasionally sampled, its probability increases, while those of the originally high-reward tokens decrease. This dynamic suggests that over-exploitation during the exploitation stage may lead to capability boundary shrinkage, whereas prolonged training into the exploration stage can promote an expansion of the reasoning capability boundary. Building upon our insights, we revisit the potential of only using relative negative gradients for prolonging training, providing a theoretical and empirical foundation for the development of more advanced reasoning capabilities. △ Less

Submitted 5 October, 2025; originally announced October 2025.

v-HUB: A Benchmark for Video Humor Understanding from Vision and Sound

Authors: Zhengpeng Shi, Yanpeng Zhao, Jianqun Zhou, Yuxuan Wang, Qinrong Cui, Wei Bi, Songchun Zhu, Bo Zhao, Zilong Zheng

Abstract: AI models capable of comprehending humor hold real-world promise -- for example, enhancing engagement in human-machine interactions. To gauge and diagnose the capacity of multimodal large language models (MLLMs) for humor understanding, we introduce v-HUB, a novel video humor understanding benchmark. v-HUB comprises a curated collection of non-verbal short videos, reflecting real-world scenarios w… ▽ More AI models capable of comprehending humor hold real-world promise -- for example, enhancing engagement in human-machine interactions. To gauge and diagnose the capacity of multimodal large language models (MLLMs) for humor understanding, we introduce v-HUB, a novel video humor understanding benchmark. v-HUB comprises a curated collection of non-verbal short videos, reflecting real-world scenarios where humor can be appreciated purely through visual cues. We pair each video clip with rich annotations to support a variety of evaluation tasks and analyses, including a novel study of environmental sound that can enhance humor. To broaden its applicability, we construct an open-ended QA task, making v-HUB readily integrable into existing video understanding task suites. We evaluate a diverse set of MLLMs, from specialized Video-LLMs to versatile OmniLLMs that can natively process audio, covering both open-source and proprietary domains. The experimental results expose the difficulties MLLMs face in comprehending humor from visual cues alone. Our findings also demonstrate that incorporating audio helps with video humor understanding, highlighting the promise of integrating richer modalities for complex video understanding tasks. △ Less

Submitted 1 June, 2026; v1 submitted 30 September, 2025; originally announced September 2025.

Comments: 24 pages, 9 figures

Identifying Dust-lane Spheroidal Galaxies in DESI Legacy Imaging Surveys Using Semi-Supervised Methods

Authors: Zhijian Luo, Jianzhen Chen, Wenxiang Pei, Hubing Xiao, Shaohua Zhang, Qifan Cui, Chenggang Shu

Abstract: Dust-lane spheroidal galaxies (DLSGs) are unique astrophysical systems that exhibit the morphology of early-type galaxies (ETGs) but are distinguished by prominent dust lanes. Recent studies propose that they form through minor mergers between ETGs and gas-rich dwarf galaxies, offering a window into the interstellar medium (ISM) of ETGs and star formation triggered by small-scale interactions. How… ▽ More Dust-lane spheroidal galaxies (DLSGs) are unique astrophysical systems that exhibit the morphology of early-type galaxies (ETGs) but are distinguished by prominent dust lanes. Recent studies propose that they form through minor mergers between ETGs and gas-rich dwarf galaxies, offering a window into the interstellar medium (ISM) of ETGs and star formation triggered by small-scale interactions. However, their rarity poses a challenge for assembling large, statistically robust samples via manual selection. To overcome this limitation, we employ GC-SWGAN, a semi-supervised learning method developed by \citet{2025ApJS..279...17L}, to systematically identify DLSGs within the DESI Legacy Imaging Surveys (DESI-LS). The methodology involves training a generative adversarial network (GAN) on unlabeled galaxy images to extract morphological features, followed by fine-tuning the model using a small dataset of labeled DLSGs. In our experiments, despite DLSGs constituting only $\sim$ 3.7\% of the test set, GC-SWGAN achieves remarkable performance, with an 87\% recall rate, 84\% accuracy, and an F1 score of 86\%, underscoring its efficacy for DLSG detection. Applying this model to $\sim$ 310,000 DESI-LS galaxies that meet the criteria $m_r < 17.0$ and $0.01 < z < 0.07$ we compile the largest catalog of DLSG candidates to date, identifying 9,482 dust-lane ETGs. A preliminary analysis reveals that these DLSGs exhibit significantly redder $g-r$ colors and higher specific star formation rates compared to non-DLSGs. This catalog enables future studies of ISM properties in ETGs and the role of minor mergers in driving star formation in the nearby universe. △ Less

Submitted 28 September, 2025; originally announced September 2025.