Showing 1–50 of 219 results for author: Pan, R

Towards Risk-free AI Agent Deployment

Authors: Yintong Huo, Rangeet Pan, Abhik Roychoudhury

Abstract: LLM-based agents are rapidly moving from research prototypes into the core business processes of organizations, but these agents pose deployment risks to security, compliance, and functionality. In this article, we argue that risk-free deployment must be grounded in the agent's trajectory: the recorded sequence of reasoning steps, tool invocations, and environmental observations. Trajectories are… ▽ More LLM-based agents are rapidly moving from research prototypes into the core business processes of organizations, but these agents pose deployment risks to security, compliance, and functionality. In this article, we argue that risk-free deployment must be grounded in the agent's trajectory: the recorded sequence of reasoning steps, tool invocations, and environmental observations. Trajectories are available for any agent, and many failures are visible only in the trajectory. To make agents deployable and sustainable, we advocate agent testing and debugging as a systematic research direction for detecting and mitigating these risks. This article begins with the challenges of testing agents, including the oracle problem, non-determinism, trajectory validation, and the absence of adequacy metrics. We then turn to debugging agents, from automated failure attribution to repair and self-evolution. We distill these directions into a practical deployment-readiness checklist covering the full deployment lifecycle. Finally, we identify open problems, i.e., formal adequacy metrics, root-cause attribution over long-horizon trajectories, and the reliability of self-evolving agents, that the community must address to enable trustworthy agent deployment. △ Less

Submitted 17 August, 2026; originally announced August 2026.

Tangent: An Empirical Study of Testing Practices for LLM-Based Agent Applications

Authors: Rangeet Pan, Tyler Stennett, Divya Sankar, Bridget McGinn, Alessandro Orso, Raju Pavuluri, Saurabh Sinha, Maja Vukovic

Abstract: Agents built on large language models (LLMs) are increasingly used to build applications that perform complex, multi-step tasks involving reasoning, tool use, and interaction with external environments. Despite rapid progress in benchmarking LLM-based agents, very few studies have attempted to understand how such systems are tested in practice. In particular, testing levels, objectives, data patte… ▽ More Agents built on large language models (LLMs) are increasingly used to build applications that perform complex, multi-step tasks involving reasoning, tool use, and interaction with external environments. Despite rapid progress in benchmarking LLM-based agents, very few studies have attempted to understand how such systems are tested in practice. In particular, testing levels, objectives, data patterns, test complexity, and validation strategies for agent applications remain underexplored. In this paper, we present an empirical study of testing practices in LLM-based agent applications using a large corpus of mined open-source projects. We construct a large-scale dataset of agent applications, tools, and tests, and manually label 2,572 test methods from 240 modules. From this analysis, we derive a taxonomy of 23 testing patterns across test fixtures, data, objectives, and assertions, and characterize tests by level (unit, module, integration). We complement this with structured interviews of 10 senior industry practitioners building agentic systems. Our results show that testing of LLM-based agent applications is dominated by narrowly scoped unit tests, with limited coverage of complex interactions, realistic scenarios, and non-functional requirements. Tests frequently rely on simplistic inputs, heavy mocking, and shallow validation, and agent-related tests exhibit low structural complexity. While industry practice places greater emphasis on non-functional testing than open-source projects, both reveal common gaps, including the lack of formal testing foundations, unclear test objectives, and challenges in generating high-quality test data. Based on these findings, we outline research directions toward more systematic and rigorous testing of agent applications, including foundations for agent testability, formalized test objectives, and fault-based testing techniques. △ Less

Submitted 8 August, 2026; originally announced August 2026.

Comments: Accepted at ASE'26

StatCite: A Large-scale Citation Network Dataset for Statistics and Data Science

Authors: Tianang Deng, Tianchen Gao, Rui Pan, Yan Zhang

Abstract: In this paper, we introduce StatCite, a large-scale citation network dataset covering publications in statistics and data science from 1981 to 2025. The dataset contains 189,101 research articles collected from 62 representative journals and provides bibliographic metadata, including title, author list, publisher, published year, abstract, keywords, and reference list. Based on the collected publi… ▽ More In this paper, we introduce StatCite, a large-scale citation network dataset covering publications in statistics and data science from 1981 to 2025. The dataset contains 189,101 research articles collected from 62 representative journals and provides bibliographic metadata, including title, author list, publisher, published year, abstract, keywords, and reference list. Based on the collected publications, we construct four complementary citation-based networks, namely the paper citation network, the co-citation network, the bibliographic coupling network, and the journal citation network. To illustrate the utility of the dataset, we present descriptive analyses of the constructed networks and investigate the community structure of the paper citation network. The results show that StatCite preserves key structural characteristics commonly observed in large-scale citation networks and captures several major research areas in statistics and data science. By integrating multiple network representations with rich textual metadata, StatCite provides a valuable resource for statistical analysis, knowledge discovery, and data-driven studies of scientific literature. △ Less

Submitted 8 August, 2026; originally announced August 2026.

Elbow-Based MoE Routing: A Training-Free Inference Time Plugin for Expert Selection

Authors: Robin Pan, Raymond Liu, Daniel Fang, Adelina Andrei, Rosa Wu

Abstract: Mixture-of-Experts (MoE) models enable model scaling while maintaining low inference-time compute by activating only a subset of experts per token. However, conventional routing relies on a fixed top-k selection, forcing the model to spend the same compute regardless of how many experts are relevant. We introduce elbow-based routing, a training-free inference-time modification that dynamically adj… ▽ More Mixture-of-Experts (MoE) models enable model scaling while maintaining low inference-time compute by activating only a subset of experts per token. However, conventional routing relies on a fixed top-k selection, forcing the model to spend the same compute regardless of how many experts are relevant. We introduce elbow-based routing, a training-free inference-time modification that dynamically adjusts the number of experts on a per-token basis. Our method examines the sorted router probability distribution and identifies an elbow point that separates high- and low-probability experts. We find that most router distributions exhibit clear inflection points suitable for this strategy, and we show both theoretically and empirically that elbow-based routing preserves expert load balance. Experiments on a state-of-the-art MoE model demonstrate an average latency reduction of 5.3% while maintaining accuracy across six benchmarks. △ Less

Submitted 4 August, 2026; originally announced August 2026.

Reusing Past Repairs Through Hierarchical Trajectory Abstraction for Coding Agents

Authors: Yisen Xu, Jiayuan Zhou, Ruiqi Pan, Tse-Hsun Chen

Abstract: Although LLM-driven repair agents can tackle complex, repository-level issues, they treat every issue independently and discard the procedural knowledge accumulated from previous repairs. We introduce STAIR, a framework that converts historical repair trajectories into hierarchical, reusable plans that can be adapted to steer future repairs. Each past trajectory is transformed into a multi-level t… ▽ More Although LLM-driven repair agents can tackle complex, repository-level issues, they treat every issue independently and discard the procedural knowledge accumulated from previous repairs. We introduce STAIR, a framework that converts historical repair trajectories into hierarchical, reusable plans that can be adapted to steer future repairs. Each past trajectory is transformed into a multi-level tree that ranges from fine-grained diagnostic actions to high-level repair strategies, encoding experience at several granularities. When a new issue arrives, STAIR selects relevant plan nodes from multiple abstraction levels, tailors them into executable, issue-specific plans, and supplies them to the agent through its prompt. On SWE-bench Verified, STAIR integrated with Lingxi reaches 81.2% Pass@1 using MiniMax M2.5 and 79.2% using GPT-5. The generated plans also generalize across agents: without any code change, they lift the Pass@1 of a structurally different agent, mini-SWE-agent v2, from 75.8% to 81.0%. Ablation experiments further show that mixing multiple abstraction levels surpasses any single level and that raw, unabstracted trajectories transfer substantially worse. △ Less

Submitted 31 July, 2026; originally announced July 2026.

Comments: 10 pages

OwlPath: Lossless Knowledge Compression for LLM Bug Repair

Authors: Bo Zhang, Ren Pan, Huan Chen, Xiang Song

Abstract: LLM-based software engineering agents are constrained by limited context windows: roughly 100K tokens must store structurally relevant code subsets to resolve bugs. Standard retrieval models treat code as plain text, forcing agents to resolve multi-hop dependencies including subclass chains, transitive callers and interface implementations through slow trial and error. We tackle this limitation… ▽ More LLM-based software engineering agents are constrained by limited context windows: roughly 100K tokens must store structurally relevant code subsets to resolve bugs. Standard retrieval models treat code as plain text, forcing agents to resolve multi-hop dependencies including subclass chains, transitive callers and interface implementations through slow trial and error. We tackle this limitation with lossless knowledge compression, encoding source code into an OWL2 ontology to answer structural queries using minimal relevant code fragments. We present OwlPath, an OWL2 reasoning layer atop CodeGraph, a widely used code intelligence platform with 500K+ GitHub stars, offering a unified CLI for structural code retrieval. Powered by tree-sitter parsing, OwlPath supports multi-language repositories (Python, JavaScript, TypeScript, Go, etc.) and encodes language-specific semantics into a unified OWL2 ontology. It adopts two complementary modules. First, a transitive-closure engine fetches all structurally linked symbols via single SPARQL property-path queries, capturing multi-hop relations missed by string matching. Second, the OWL Software Knowledge Map (OWL-SKM) precomputes a compact 3KB summary with module trees, core APIs and issue-related symbols, directing agents to target modules in the first query. Evaluated on 18 SWE-bench Pro instances, OwlPath obtains a 68.4% strict-apply rate versus 66.7% for the CodeGraph baseline, cutting token usage by 28.8% and runtime by 39.5%. In offline retrieval tests over 67 instances, OwlPath improves recall 2.06 times (0.464 vs 0.226) and reaches 88.1% hit rate compared to CodeGraph's 59.7%. On a 37-question structural retrieval benchmark, recall rises from 4.4% to 28.8%, with 69-80% accuracy on transitive caller and interface tasks. △ Less

Submitted 28 July, 2026; originally announced July 2026.

SuperPass: Fast-Tracking Blocking Threads to Mitigate Priority Inversion on Mobile Devices

Authors: Lei Li, Yu Liang, Riwei Pan, Youcheng Sun, Nan Guan, Tei-Wei Kuo, Chun Jason Xue

Abstract: Priority inversion occurs when a high-priority thread is delayed by a lower-priority one. Although well studied in real-time systems, its impact in general-purpose OSes (e.g., Android) remains underexplored. On Android, we find that priority inversions happen frequently and can delay latency-critical threads, degrading user experience. For example, the foreground app's UI thread is frequently bloc… ▽ More Priority inversion occurs when a high-priority thread is delayed by a lower-priority one. Although well studied in real-time systems, its impact in general-purpose OSes (e.g., Android) remains underexplored. On Android, we find that priority inversions happen frequently and can delay latency-critical threads, degrading user experience. For example, the foreground app's UI thread is frequently blocked by low-priority threads, with blocking durations of up to 210 ms, enough to cause dropped frames. Existing solutions designed for real-time systems fail to eliminate long priority-inversion blockings on latency-critical threads and may introduce high overhead on Android. To solve this problem, we uncover two insights on Android: 1) long blockings are mainly due to the accumulated CPU waiting time of low-priority blocking threads rather than their critical-section latency; and 2) although latency-critical threads can be blocked by many concurrent readers, tracking a limited number of them is sufficient to achieve good responsiveness with low overhead in most cases. Guided by these insights, we propose SuperPass, a lightweight kernel mechanism that mitigates priority inversion by fast-track scheduling of low-priority threads blocking latency-critical threads. It introduces a scheduler fast track that grants immediate CPU access to threads blocking latency-critical threads, and employs a lock-level detector that effectively identifies most such blocking threads. We evaluate SuperPass on a Google Pixel 8 smartphone. Taking UI thread as a case study, SuperPass decreases the 99.9th-percentile blocking duration by 72.0% and blocking count by 47.7% on average compared to the default scheduler, and reduces janky frames by 29.2% with a system-wide CPU overhead of only 0.74%. SuperPass also outperforms existing approaches including priority inheritance, real-time UI promotion, and Proxy Execution. △ Less

Submitted 21 July, 2026; v1 submitted 20 July, 2026; originally announced July 2026.

PraMem: Practice-derived Experiential Memory for Long-horizon Behavior Prediction

Authors: Zhuoqun Li, Boxi Cao, Jiawei Chen, Hanshu Zhou, Ruoxi Xu, Guiping Jiang, Ruotong Pan, Tingting Gao, Han Li, Xiangyu Wu, Hongyu Lin, Yaojie Lu, Xianpei Han, Le Sun

Abstract: Long-horizon behavior prediction aims to infer a user's next action based on a lengthy historical sequence, playing a crucial role in artificial intelligence field. The rise of large language models (LLMs) offers a promising direction for sequential behavior prediction, yet LLMs struggle with latent behavioral pattern induction and model-intrinsic cognitive biases when tackling long-horizon behavi… ▽ More Long-horizon behavior prediction aims to infer a user's next action based on a lengthy historical sequence, playing a crucial role in artificial intelligence field. The rise of large language models (LLMs) offers a promising direction for sequential behavior prediction, yet LLMs struggle with latent behavioral pattern induction and model-intrinsic cognitive biases when tackling long-horizon behavior prediction. Prior memory management methods follow a context-compression paradigm that attempts to address this task by alleviating the historical sequence burden, yet fail to resolve the core challenges. In this paper, we advocate a paradigm shift that reframes the lengthy historical sequence from a burden into a valuable resource to be exploited, and accordingly propose PraMem, which conducts beforehand practice over the lengthy historical sequence to build an experiential memory, thereby serving as the assisted input for accurate long-horizon behavior prediction. Extensive experiments across diverse tasks demonstrate that PraMem achieves superior performance than prior methods, and more in-depth analyses provide valuable insights into the mechanism and evolution of the experiential memory. Code: https://github.com/icip-cas/PraMem. △ Less

Submitted 2 July, 2026; originally announced July 2026.

An LLM-Powered Semantic Alignment Framework for Journal Recommendation

Authors: Yanglin Yan, Zicheng Xie, Tianchen Gao, Rui Pan, Hansheng Wang

Abstract: Journal recommendation is an important task in scholarly information systems. Existing approaches typically rely on supervised learning models, manually engineered features, or historical interaction data, which may limit their generalizability and interpretability. We propose an LLM-powered semantic alignment framework that formulates journal recommendation as a semantic matching problem between… ▽ More Journal recommendation is an important task in scholarly information systems. Existing approaches typically rely on supervised learning models, manually engineered features, or historical interaction data, which may limit their generalizability and interpretability. We propose an LLM-powered semantic alignment framework that formulates journal recommendation as a semantic matching problem between manuscript content and journal scope descriptions. The framework enables large language models (LLMs) to infer journal suitability directly from article titles, abstracts, keywords, and candidate journal information without task-specific training. Experiments are conducted using DeepSeek-V3 on a dataset of 23,609 articles from 49 journals in statistics and related fields. The proposed framework achieves Top-3, Top-5, and Top-10 accuracies of 40.23\%, 53.67\%, and 70.05\%, respectively. Additional analyses show that incorporating reference information generally improves recommendation performance and that recommendations remain highly stable across repeated runs, with an average Top-5 Jaccard similarity of 84\%. The framework also generates interpretable reasoning outputs that provide insights into the recommendation process. These findings demonstrate the potential of LLMs as a training-free and scalable paradigm for journal recommendation and scholarly decision support. △ Less

Submitted 26 June, 2026; originally announced June 2026.

DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence

Authors: DeepSeek-AI, Anyi Xu, Bangcai Lin, Bing Xue, Bingxuan Wang, Bingzheng Xu, Bochao Wu, Bowei Zhang, Chaofan Lin, Chen Dong, Chenchen Ling, Chengda Lu, Chenggang Zhao, Chengqi Deng, Chengyu Hou, Chenhao Xu, Chenze Shao, Chong Ruan, Conner Sun, Damai Dai, Daya Guo, Dejian Yang, Deli Chen, Donghao Li, Dongjie Ji , et al. (294 additional authors not shown)

Abstract: We present a preview version of DeepSeek-V4 series, including two strong Mixture-of-Experts (MoE) language models -- DeepSeek-V4-Pro with 1.6T parameters (49B activated) and DeepSeek-V4-Flash with 284B parameters (13B activated) -- both supporting a context length of one million tokens. DeepSeek-V4 series incorporate several key upgrades in architecture and optimization: (1) a hybrid attention arc… ▽ More We present a preview version of DeepSeek-V4 series, including two strong Mixture-of-Experts (MoE) language models -- DeepSeek-V4-Pro with 1.6T parameters (49B activated) and DeepSeek-V4-Flash with 284B parameters (13B activated) -- both supporting a context length of one million tokens. DeepSeek-V4 series incorporate several key upgrades in architecture and optimization: (1) a hybrid attention architecture that combines Compressed Sparse Attention (CSA) and Heavily Compressed Attention (HCA) to improve long-context efficiency; (2) Manifold-Constrained Hyper-Connections (mHC) that enhance conventional residual connections; (3) and the Muon optimizer for faster convergence and greater training stability. We pre-train both models on more than 32T diverse and high-quality tokens, followed by a comprehensive post-training pipeline that unlocks and further enhances their capabilities. DeepSeek-V4-Pro-Max, the maximum reasoning effort mode of DeepSeek-V4-Pro, redefines the state-of-the-art for open models, outperforming its predecessors in core tasks. Meanwhile, DeepSeek-V4 series are highly efficient in long-context scenarios. In the one-million-token context setting, DeepSeek-V4-Pro requires only 27% of single-token inference FLOPs and 10% of KV cache compared with DeepSeek-V3.2. This enables us to routinely support one-million-token contexts, thereby making long-horizon tasks and further test-time scaling more feasible. The model checkpoints are available at https://huggingface.co/collections/deepseek-ai/deepseek-v4. △ Less

Submitted 26 April, 2026; originally announced June 2026.

Diffusion-Proof: Recipe for Formal Theorem Proving Beyond Auto-Regressive Generation

Authors: Ruida Wang, Rui Pan, Pengcheng Wang, Shizhe Diao, Tong Zhang

Abstract: Enhancing the formal math reasoning capabilities of Large Language Models (LLMs) has become a key focus in both mathematical and computer science communities in recent years. While significant progress has been made in using state-of-the-art Auto-Regressive (AR) LLMs for formal theorem proving, these models suffer from inherent limitations. Their next-token prediction generation methods may yield… ▽ More Enhancing the formal math reasoning capabilities of Large Language Models (LLMs) has become a key focus in both mathematical and computer science communities in recent years. While significant progress has been made in using state-of-the-art Auto-Regressive (AR) LLMs for formal theorem proving, these models suffer from inherent limitations. Their next-token prediction generation methods may yield suboptimal performance due to the challenges of long-range coherence and the compounding of errors over long sequences. Recent advancements in diffusion LLMs (dLLMs), which generate text through iterative denoising of a multi-token block, offer a promising alternative. However, the application of dLLMs to formal mathematics, where maintaining long-range coherence is critical, remains largely understudied. To address the challenges above, we propose **Diffusion-Proof**, to the best of our knowledge, the first framework to train and apply dLLMs for formal theorem proving. Our frameworks contain training and inference methods for two models. The first one is *dLLM-Prover-7B*, which performs whole-proof writing with long-range coherent tactic usage. The second one is *dLLM-Corrector-7B*, which is a novel large block diffusion-based correction model. It leverages the in-filling capabilities of dLLMs to perform local proof correction using bi-directional information. Extensive experiments demonstrate that **Diffusion-Proof** relatively significantly outperforms the AR LLM baseline trained under the same dataset. **Diffusion-Proof** achieves an absolute improvement of **1.61%** on ProofNet-Test and **6.14%** on MiniF2F-Test benchmarks compare to the baseline. Notably, **Diffusion-Proof** successfully resolves one IMO problem that more advanced thinking model DeepSeek-Prover-V2-7B could not solve, showcasing the unique advantage of dLLMs in formal theorem proving. △ Less

Submitted 17 June, 2026; originally announced June 2026.

The Multipath Reliable Connection (MRC) Transport

Authors: Rip Sohan, Eric Spada, Eric Davis, Mark Handley, Idan Burstein, Tony Hurson, Jithin Jose, Vivek Kashyap, Rong Pan, Sayantan Sur, Sreevatsa Anantharamu, Aviv Barnea, Adrian Caulfield, Elazar Cohen, Elliot Edmunds, Yamin Friedman, Mahdieh Ghazi, Murali Guramali, Torsten Hoefler, Vipin Jain, Abdul Kabbani, Noam Katz, Yanfang Le, Charlie Mbariky, Guglielmo Morandin , et al. (14 additional authors not shown)

Abstract: MRC is an open, production-grade transport designed for large-scale AI/ML training over best-effort Ethernet. It extends RoCEv2 with explicit, composable primitives for per-packet multipath and sender-based congestion control, decouples packet delivery from semantic processing, adds multiple new capabilities for accelerated packet-loss recovery and adds resilience against port and path failures. T… ▽ More MRC is an open, production-grade transport designed for large-scale AI/ML training over best-effort Ethernet. It extends RoCEv2 with explicit, composable primitives for per-packet multipath and sender-based congestion control, decouples packet delivery from semantic processing, adds multiple new capabilities for accelerated packet-loss recovery and adds resilience against port and path failures. This paper presents MRC and details its core capabilities and mechanisms. △ Less

Submitted 16 June, 2026; originally announced June 2026.

Sakura: An Approach for Generating Complex Tests from Natural Language Test Descriptions

Authors: Tyler Stennett, Rangeet Pan, Bridget McGinn, Alessandro Orso, Saurabh Sinha

Abstract: Research on automating software testing has spanned several decades. Most existing approaches generate unit tests for individual methods, validate isolated API endpoints, or target user interface (UI) layers, with non-API and non-UI generators typically exercising only a single focal method. Recent empirical evidence shows a substantial gap between such generated tests and developer-written ones,… ▽ More Research on automating software testing has spanned several decades. Most existing approaches generate unit tests for individual methods, validate isolated API endpoints, or target user interface (UI) layers, with non-API and non-UI generators typically exercising only a single focal method. Recent empirical evidence shows a substantial gap between such generated tests and developer-written ones, which often span several focal classes and methods, involve multi-step call sequences, and contain chained assertions, all characteristics that current approaches fail to capture. To address this gap, we propose generating tests from natural language (NL) descriptions of developer intent, an expressive and accessible medium for specifying complex test scenarios. We present Sakura, the first agent-based framework for generating structurally complex tests from NL descriptions. Sakura decomposes NL descriptions into structured blocks and processes them with a multi-agent system: a localization agent grounds test steps in concrete application code via static analysis, a composition agent synthesizes compilable test code and iteratively refines it using execution feedback, and a supervisor agent coordinates their interactions. To evaluate Sakura, we curate a novel dataset of NL test descriptions at three levels of abstraction, reflecting different end-user personas, systematically derived from developer-written tests in Apache Commons projects. Across 20 applications and 1,464 test scenarios, Sakura substantially outperforms off-the-shelf agentic tools such as Gemini CLI instantiated with multiple LLMs, achieving 50-78% higher test compilability and 38-66% higher coverage overlap with ground-truth tests using the same models. Moreover, Sakura paired with small open-source models such as Devstral Small 2 and Qwen3-Coder outperforms Gemini CLI using large proprietary models, at lower cost. △ Less

Submitted 13 August, 2026; v1 submitted 30 May, 2026; originally announced June 2026.

Comments: 25 pages, 11 figures. Accepted to ISSTA 2026

ICCU: In-Context Continual Unlearning via Pattern-Induced Refusal Rules

Authors: Ruihao Pan, Suhang Wang

Abstract: Machine unlearning aims to remove the influence of specific data from trained language models. In real-world deployments, unlearning requests often arrive sequentially, which challenges existing fine-tuning-based methods: fine-tuning each request is costly, accumulates utility loss, and may cause cross-request interference. To address these issues, we propose ICCU (In-Context Continual Unlearning)… ▽ More Machine unlearning aims to remove the influence of specific data from trained language models. In real-world deployments, unlearning requests often arrive sequentially, which challenges existing fine-tuning-based methods: fine-tuning each request is costly, accumulates utility loss, and may cause cross-request interference. To address these issues, we propose ICCU (In-Context Continual Unlearning), an in-context continual unlearning framework that induces readable refusal rules from unlearning datasets and applies them at inference time either as a filter or via the system prompt, without modifying model parameters. Because rules are accumulated as an order-independent union, ICCU is compositional and free of cross-request interference, and the original forget-set data can be discarded after rule induction. Extensive experiments show that ICCU effectively suppresses target knowledge while preserving utility, scales across sequential requests, and remains robust to paraphrased and cross-lingual queries. △ Less

Submitted 26 May, 2026; originally announced May 2026.

MetaphorVU: Towards Metaphorical Video Understanding

Authors: Zhuoqun Li, Boxi Cao, Guiping Jiang, Fangrui Lv, Ruotong Pan, Jianan Wang, Xiangyu Wu, Hongyu Lin, Yaojie Lu, Yong Du, Ruyin Jia, Liyan, Tingting Gao, Han Li, Xianpei Han, Le Sun

Abstract: Metaphorical videos are prevalent across various real-world scenarios to convey complex ideas, and understanding them typically requires high-order cognitive capabilities. The lack of systematic studies on metaphorical video understanding not only constrains the real-world applicability of MLLMs but also impedes the thorough assessment of their high-order cognitive capabilities. To bridge this gap… ▽ More Metaphorical videos are prevalent across various real-world scenarios to convey complex ideas, and understanding them typically requires high-order cognitive capabilities. The lack of systematic studies on metaphorical video understanding not only constrains the real-world applicability of MLLMs but also impedes the thorough assessment of their high-order cognitive capabilities. To bridge this gap, we propose MetaphorVU-Bench, the first systematic and comprehensive benchmark dedicated to metaphorical video understanding. Through experiments, we find current MLLMs struggle with accurate metaphorical video understanding, lagging far behind human level, primarily due to defective cross-domain mapping. Motivated by this finding, we construct a metaphor knowledge graph as mapping augmentation and propose MetaphorBoost, an inference-time enhancement framework achieving consistent performance improvement. Our benchmark, analysis, and method provide useful insights and a foundation for future research on advancing MLLMs. △ Less

Submitted 25 May, 2026; originally announced May 2026.

Comments: ICML 2026 spotlight

GoLongRL: Capability-Oriented Long Context Reinforcement Learning with Multitask Alignment

Authors: Minxuan Lv, Tiehua Mei, Tanlong Du, Junmin Chen, Zhenpeng Su, Ziyang Chen, Ziqi Wang, Zhennan Wu, Ruotong Pan, jian Liang, Ruiming Tang, Han Li

Abstract: We present GoLongRL, a fully open-source, capability-oriented post-training recipe for long-context reinforcement learning with verifiable rewards (RLVR). Existing long-context RL methods often treat data construction as a matter of designing increasingly complex retrieval paths, leading to homogeneous task coverage and reward formulations that inadequately reflect practical long-context requireme… ▽ More We present GoLongRL, a fully open-source, capability-oriented post-training recipe for long-context reinforcement learning with verifiable rewards (RLVR). Existing long-context RL methods often treat data construction as a matter of designing increasingly complex retrieval paths, leading to homogeneous task coverage and reward formulations that inadequately reflect practical long-context requirements. Our work offers two contributions. (1) Capability-oriented data construction with full open release. We openly release a dataset of 23K RLVR samples, the complete construction pipeline, and all training code. Guided by a taxonomy of long-context capabilities, the dataset spans 9 task types, each paired with its natural evaluation metric. It comprises curated open-source samples from established corpora and synthetic samples whose QA pairs are generated from real source documents such as books, academic papers, and multi-turn dialogues. Under the same vanilla GRPO setup, our dataset alone outperforms the closed-source QwenLong-L1.5 dataset. Moreover, our Qwen3-30B-A3B model trained on this data delivers long-context performance comparable to DeepSeek-R1-0528 and Qwen3-235B-A22B-Thinking-2507, suggesting that broader coverage and greater reward diversity substantially benefit long-context capability improvement. (2) TMN-Reweight for heterogeneous multitask optimization. To address optimization challenges from heterogeneous rewards, we propose TMN-Reweight, which combines task-level mean normalization for cross-task reward scale alignment with difficulty-adaptive weighting for more reliable advantage estimation. TMN-Reweight further improves average performance over vanilla GRPO, with general capabilities preserved or improved across reported evaluations. △ Less

Submitted 19 May, 2026; originally announced May 2026.

Supporting System Testing with a Multi-Agent LLM-based Framework for Knowledge Graph Extraction: A Case Study with Ethernet Switch Systems

Authors: Rongqi Pan, Mahboubeh Dadkhah, Jean Baptiste Minani, Hussein Al Osman, Lionel Briand, Haiwei Dong

Abstract: Technical documents contain rich domain knowledge for automating downstream tasks such as system testing. While this paper focuses on Ethernet switch configuration manuals (ESCMs), we propose a general framework that can be adapted to different industrial contexts. ESCMs provide valuable domain knowledge for Ethernet switch testing, but their semi-structured format, implicit step attributes, and c… ▽ More Technical documents contain rich domain knowledge for automating downstream tasks such as system testing. While this paper focuses on Ethernet switch configuration manuals (ESCMs), we propose a general framework that can be adapted to different industrial contexts. ESCMs provide valuable domain knowledge for Ethernet switch testing, but their semi-structured format, implicit step attributes, and complex section dependencies make them difficult to directly leverage for test automation. To address this, we generate knowledge graphs (KGs) that capture configuration knowledge from ESCM in a structured form. We propose a multi-agent LLM-based framework that extracts, evaluates, and improves KGs from ESCMs using a fine-grained KG schema and an iterative Extract-Evaluate-Improve (EEI) loop. Our evaluation on 50 real-world ESCMs shows that our framework achieves high extraction correctness using the original prompts, with average correctness scores ranging from 0.97 to 0.99 across three extraction tasks. For challenging ESCMs, the EEI loop further improves correctness through manual-specific prompt refinement. Moreover, the LLM judgments and human evaluations show substantial agreement, with Cohen's kappa of at least 0.72 across all extraction tasks. Finally, feedback from industry testers indicates that the generated KGs can support the generation of useful and correct test case specifications (TCSs) for downstream testing. △ Less

Submitted 18 May, 2026; originally announced May 2026.

Code as Agent Harness

Authors: Xuying Ning, Katherine Tieu, Dongqi Fu, Tianxin Wei, Zihao Li, Yuanchen Bei, Jiaru Zou, Mengting Ai, Zhining Liu, Ting-Wei Li, Lingjie Chen, Yanjun Zhao, Ke Yang, Bingxuan Li, Cheng Qian, Gaotang Li, Xiao Lin, Zhichen Zeng, Ruizhong Qiu, Sirui Chen, Yifan Sun, Xiyuan Yang, Ruida Wang, Rui Pan, Chenyuan Yang , et al. (17 additional authors not shown)

Abstract: Recent large language models (LLMs) have demonstrated strong capabilities in understanding and generating code, from competitive programming to repository-level software engineering. In emerging agentic systems, code is no longer only a target output. It increasingly serves as an operational substrate for agent reasoning, acting, environment modeling, and execution-based verification. We frame thi… ▽ More Recent large language models (LLMs) have demonstrated strong capabilities in understanding and generating code, from competitive programming to repository-level software engineering. In emerging agentic systems, code is no longer only a target output. It increasingly serves as an operational substrate for agent reasoning, acting, environment modeling, and execution-based verification. We frame this shift through the lens of agent harnesses and introduce code as agent harness: a unified view that centers code as the basis for agent infrastructure. To systematically study this perspective, we organize the survey around three connected layers. First, we study the harness interface, where code connects agents to reasoning, action, and environment modeling. Second, we examine harness mechanisms: planning, memory, and tool use for long-horizon execution, together with feedback-driven control and optimization that make harness reliable and adaptive. Third, we discuss scaling the harness from single-agent systems to multi-agent settings, where shared code artifacts support multi-agent coordination, review, and verification. Across these layers, we summarize representative methods and practical applications of code as agent harness, spanning coding assistants, GUI/OS automation, embodied agents, scientific discovery, personalization and recommendation, DevOps, and enterprise workflows. We further outline open challenges for harness engineering, including evaluation beyond final task success, verification under incomplete feedback, regression-free harness improvement, consistent shared state across multiple agents, human oversight for safety-critical actions, and extensions to multimodal environments. By centering code as the harness of agentic AI, this survey provides a unified roadmap toward executable, verifiable, and stateful AI agent systems. △ Less

Submitted 18 May, 2026; originally announced May 2026.

Comments: GitHub: https://github.com/YennNing/Awesome-Code-as-Agent-Harness-Papers

Slipstream: Trajectory-Grounded Compaction Validation for Long-Horizon Agents

Authors: Zhuofu Chen, Rui Pan, Yinwei Dai, Ravi Netravali

Abstract: To cope with the large contexts that long-horizon LLM agents produce, modern frameworks increasingly rely on compaction -- invoking an LLM to rewrite the accumulated trajectory into a shorter summary that the agent resumes from. Today, compaction runs synchronously on the critical path of agent execution but this can unpredictably degrade accuracy due to a structural validation gap: the compactor… ▽ More To cope with the large contexts that long-horizon LLM agents produce, modern frameworks increasingly rely on compaction -- invoking an LLM to rewrite the accumulated trajectory into a shorter summary that the agent resumes from. Today, compaction runs synchronously on the critical path of agent execution but this can unpredictably degrade accuracy due to a structural validation gap: the compactor must condense context but is fundamentally unaware of precisely what information the agent will need later. Further, because post-compaction agent steps are conditioned on the new summary, targeted validation criteria do not exist and errors silently propagate through coherent but incorrect behavior. Our key insight is that asynchronous compaction efficiently addresses this gap: by running the compactor in parallel with continued agent execution on the original context, the candidate summary and the agent's next steps are generated independently from the same pre-compaction state, yielding a validation signal independent of the summary itself. We build Slipstream, a trajectory-grounded compaction system that uses a judge to validate the candidate summary against the agent's continued reasoning, checking that it preserves both the agent's forward intent and the key facts and constraints it depends on. Across long-horizon coding (SWE-bench Verified) and web-browsing (BrowseComp) workloads, Slipstream improves task accuracy by up to 8.8 percentage points while reducing end-to-end latency by up to 39.7%. △ Less

Submitted 8 May, 2026; originally announced May 2026.

Comments: 9 pages (16 pages counting references, appendix), 6 figures, 2 tables

ACM Class: I.2.0

From Diffusion to Rectified Flow: Rethinking Text-Based Segmentation

Authors: Zishen Qu, Xuesong Li, Haijian Gu, Hongwei Kang, Quan Meng, Tianrui Niu, Xin Yang, Ruidong Pan

Abstract: Text-based image segmentation aims to delineate object boundaries within an image from text prompts, offering higher flexibility and broader application scope compared to traditional fixed-category segmentation tasks. Recent studies have shown that diffusion models (e.g., Stable Diffusion) can provide rich multimodal semantic features, leading to studies of using diffusion models as feature extrac… ▽ More Text-based image segmentation aims to delineate object boundaries within an image from text prompts, offering higher flexibility and broader application scope compared to traditional fixed-category segmentation tasks. Recent studies have shown that diffusion models (e.g., Stable Diffusion) can provide rich multimodal semantic features, leading to studies of using diffusion models as feature extractors for segmentation tasks. Such methods, however, inherit the generative natures of diffusion models that are harmful to discriminative segmentation tasks. In response, we propose RLFSeg, a novel framework that leverages Rectified Flow to learn direct mapping from the image to the segmentation mask within the latent space. The model is thus freed from the noise-denoise process and the need to optimize the time step of diffusion models, resulting in substantially better performance than previous diffusion-based methods, especially on zero-shot scenarios. By introducing label refinement and an Adaptive One-Step Sampling strategy, the model achieves higher accuracy even on a single inference step. The framework redirects a pretrained generative model to the discriminative segmentation task with zero modification to model structure, thus reveals promising application potential and significant research value. △ Less

Submitted 6 May, 2026; originally announced May 2026.

Comments: Accepted at ICMR 2026

Resilient AI Supercomputer Networking using MRC and SRv6

Authors: Joao Araujo, Alex Chow, Mark Handley, Ryder Lewis, Christoph Paasch, Jitendra Padhye, Michael Papamichael, Greg Steinbrecher, Amin Tootoonchian, Lihua Yuan, S. Anantharamu, Abhishek Dosi, Mohit Garg, Mahdieh Ghazi, Torsten Hoefler, Deepal Jayasinghe, Jithin Jose, Abdul Kabbani, Guohan Lu, Yang Wang, K. Doddapaneni, Murali Garimella, Vipin Jain, Yanfang Le, H. Nagulapalli , et al. (25 additional authors not shown)

Abstract: Tail latency dominates the performance of synchronous pretraining jobs when running at very large scales. We describe a three-pronged approach: (1) a new RDMA-based transport protocol, MRC, sprays across many paths and actively load-balances between them, eliminating the issue of flow collisions (2) the use of multi-plane Clos topologies to get the benefits of high switch radix and redundancy, all… ▽ More Tail latency dominates the performance of synchronous pretraining jobs when running at very large scales. We describe a three-pronged approach: (1) a new RDMA-based transport protocol, MRC, sprays across many paths and actively load-balances between them, eliminating the issue of flow collisions (2) the use of multi-plane Clos topologies to get the benefits of high switch radix and redundancy, allowing training clusters well over 100K GPUs to be built as two-tier topologies while increasing physical redundancy, and (3) the use of static source-routing using SRv6 to allow MRC the freedom to bypass failures by itself. We describe our experiences running MRC and static SRv6 routing in production in OpenAI and Microsoft's largest training clusters, where it has been used to train the latest frontier models. We demonstrate how MRC allows AI training jobs to ride out many network failures that previously would have interrupted training. △ Less

Submitted 5 May, 2026; originally announced May 2026.

Comments: 18 pages, 22 figures

ACM Class: C.2.2; I.2

Recursive Multi-Agent Systems

Authors: Jiaru Zou, Rui Pan, Ruizhong Qiu, Pan Lu, Shizhe Diao, Jindong Jiang, Hanghang Tong, Tong Zhang, Markus J. Buehler, Jingrui He, James Zou

Abstract: Recursive or looped language models have recently emerged as a new scaling axis by iteratively refining the same model computation over latent states to deepen reasoning. We extend such scaling principle from a single model to multi-agent systems, and ask: Can agent collaboration itself be scaled through recursion? To this end, we introduce RecursiveMAS, a recursive multi-agent framework that cast… ▽ More Recursive or looped language models have recently emerged as a new scaling axis by iteratively refining the same model computation over latent states to deepen reasoning. We extend such scaling principle from a single model to multi-agent systems, and ask: Can agent collaboration itself be scaled through recursion? To this end, we introduce RecursiveMAS, a recursive multi-agent framework that casts the entire system as a unified latent-space recursive computation. RecursiveMAS connects heterogeneous agents as a collaboration loop through the lightweight RecursiveLink module, enabling in-distribution latent thoughts generation and cross-agent latent state transfer. To optimize our framework, we develop an inner-outer loop learning algorithm for iterative whole-system co-optimization through shared gradient-based credit assignment across recursion rounds. Theoretical analyses of runtime complexity and learning dynamics establish that RecursiveMAS is more efficient than standard text-based MAS and maintains stable gradients during recursive training. Empirically, we instantiate RecursiveMAS under 4 representative agent collaboration patterns and evaluate across 9 benchmarks spanning mathematics, science, medicine, search, and code generation. In comparison with advanced single/multi-agent and recursive computation baselines, RecursiveMAS consistently delivers an average accuracy improvement of 8.3%, together with 1.2$\times$-2.4$\times$ end-to-end inference speedup, and 34.6%-75.6% token usage reduction. Code and Data are provided in https://recursivemas.github.io. △ Less

Submitted 12 July, 2026; v1 submitted 28 April, 2026; originally announced April 2026.

Comments: Project Website: https://recursivemas.github.io

StoSignSGD: Unbiased Structural Stochasticity Fixes SignSGD for Training Large Language Models

Authors: Dingzhi Yu, Rui Pan, Yuxing Liu, Tong Zhang

Abstract: Sign-based optimization algorithms, such as SignSGD, have garnered significant attention for their remarkable performance in distributed learning and training large foundation models. Despite their empirical superiority, SignSGD is known to diverge on non-smooth objectives, which are ubiquitous in modern machine learning due to ReLUs, max-pools, and mixture-of-experts. To overcome this fundamental… ▽ More Sign-based optimization algorithms, such as SignSGD, have garnered significant attention for their remarkable performance in distributed learning and training large foundation models. Despite their empirical superiority, SignSGD is known to diverge on non-smooth objectives, which are ubiquitous in modern machine learning due to ReLUs, max-pools, and mixture-of-experts. To overcome this fundamental limitation, we propose \textbf{StoSignSGD}, an algorithm that injects structural stochasticity into the sign operator while maintaining an unbiased update step. In the regime of (online) convex optimization, our theoretical analysis shows that StoSignSGD rigorously resolves the non-convergence issues of SignSGD, achieving a sharp convergence rate matching the lower bound. For the more challenging non-convex non-smooth optimization, we introduce generalized stationary measures that encompass prior definitions, proving that StoSignSGD improves upon the best-known complexity bounds by dimensional factors. Empirically, StoSignSGD exhibits robust stability and superior efficiency across diverse large language model (LLM) training regimes. Notably, in low-precision FP8 pretraining -- a setting where AdamW fails catastrophically -- StoSignSGD remains highly stable and yields a remarkable 1.44$\times$ to 2.14$\times$ speedup relative to established baselines. Furthermore, when fine-tuning 7B LLMs on mathematical reasoning tasks, StoSignSGD delivers substantial performance gains over both AdamW and SignSGD. Finally, to dissect the mechanisms driving its success, we develop a sign conversion framework capable of transforming any general optimizer into its unbiased, sign-based counterpart. Utilizing this framework, we deconstruct the core components of StoSignSGD and present a comprehensive ablation study to empirically validate our algorithmic design choices. △ Less

Submitted 16 April, 2026; originally announced April 2026.

AgentSPEX: An Agent SPecification and EXecution Language

Authors: Pengcheng Wang, Jerry Huang, Jiarui Yao, Rui Pan, Peizhi Niu, Yaowenqi Liu, Ruida Wang, Renhao Lu, Yuwei Guo, Tong Zhang

Abstract: Language-model agent systems commonly rely on reactive prompting, in which a single instruction guides the model through an open-ended sequence of reasoning and tool-use steps, leaving control flow and intermediate state implicit and making agent behavior potentially difficult to control. Orchestration frameworks such as LangGraph, DSPy, and CrewAI impose greater structure through explicit workflo… ▽ More Language-model agent systems commonly rely on reactive prompting, in which a single instruction guides the model through an open-ended sequence of reasoning and tool-use steps, leaving control flow and intermediate state implicit and making agent behavior potentially difficult to control. Orchestration frameworks such as LangGraph, DSPy, and CrewAI impose greater structure through explicit workflow definitions, but tightly couple workflow logic with Python, making agents difficult to maintain and modify. In this paper, we introduce AgentSPEX, an Agent SPecification and EXecution Language for specifying LLM-agent workflows with explicit control flow and modular structure, along with a customizable agent harness. AgentSPEX supports typed steps, branching and loops, parallel execution, reusable submodules, and explicit state management, and these workflows execute within an agent harness that provides tool access, a sandboxed virtual environment, and support for checkpointing, verification, and logging. Furthermore, we provide a visual editor with synchronized graph and workflow views for authoring and inspection. We include ready-to-use agents for deep research and scientific research, and we evaluate AgentSPEX on 7 benchmarks. Finally, we show through a user study that AgentSPEX provides a more interpretable and accessible workflow-authoring paradigm than a popular existing agent framework. △ Less

Submitted 14 April, 2026; originally announced April 2026.

Towards Real-world Human Behavior Simulation: Benchmarking Large Language Models on Long-horizon, Cross-scenario, Heterogeneous Behavior Traces

Authors: Jiawei Chen, Ruoxi Xu, Boxi Cao, Ruotong Pan, Yunfei Zhang, Yifei Hu, Yong Du, Tingting Gao, Yaojie Lu, Yingfei Sun, Xianpei Han, Le Sun, Xiangyu Wu, Hongyu Lin

Abstract: The emergence of Large Language Models (LLMs) has illuminated the potential for a general-purpose user simulator. However, existing benchmarks remain constrained to isolated scenarios, narrow action spaces, or synthetic data, failing to capture the holistic nature of authentic human behavior. To bridge this gap, we introduce OmniBehavior, the first user simulation benchmark constructed entirely fr… ▽ More The emergence of Large Language Models (LLMs) has illuminated the potential for a general-purpose user simulator. However, existing benchmarks remain constrained to isolated scenarios, narrow action spaces, or synthetic data, failing to capture the holistic nature of authentic human behavior. To bridge this gap, we introduce OmniBehavior, the first user simulation benchmark constructed entirely from real-world data, integrating long-horizon, cross-scenario, and heterogeneous behavioral patterns into a unified framework. Based on this benchmark, we first provide empirical evidence that previous datasets with isolated scenarios suffer from tunnel vision, whereas real-world decision-making relies on long-term, cross-scenario causal chains. Extensive evaluations of state-of-the-art LLMs reveal that current models struggle to accurately simulate these complex behaviors, with performance plateauing even as context windows expand. Crucially, a systematic comparison between simulated and authentic behaviors uncovers a fundamental structural bias: LLMs tend to converge toward a positive average person, exhibiting hyper-activity, persona homogenization, and a utopian bias. This results in the loss of individual differences and long-tail behaviors, highlighting critical directions for future high-fidelity simulation research. △ Less

Submitted 21 May, 2026; v1 submitted 9 April, 2026; originally announced April 2026.

Comments: Project page: https://OmniBehavior.github.io

Enabling Deterministic User-Level Interrupts in Real-Time Processors via Hardware Extension

Authors: Hongbin Yang, Huanle Zhang, Runyu Pan

Abstract: The growing complexity of real-time embedded systems demands strong isolation of software components into separate protection domains to reduce attack surfaces and limit fault propagation. However, application-supplied device interrupt handlers -- even untrusted -- have to remain in the kernel to minimize interrupt latency, undermining security and burdening manual certifications. Current hardware… ▽ More The growing complexity of real-time embedded systems demands strong isolation of software components into separate protection domains to reduce attack surfaces and limit fault propagation. However, application-supplied device interrupt handlers -- even untrusted -- have to remain in the kernel to minimize interrupt latency, undermining security and burdening manual certifications. Current hardware extensions accelerate interrupts only when the target protection domain is scheduled by the kernel; consequently, they are limited to improving average-case performance but not worst-case latency, and do not meet the requirements of critical real-time applications such as autonomous vehicles or robots. To overcome this limitation, we propose a novel hardware extension that enables direct, deterministic switching to the appropriate protection domain upon user-level interrupt arrival -- without kernel intervention -- even when that domain is dormant. Our hardware extension reduces worst-case latency by more than 50x with a 19% increase in core area (2% of total die area) and 4.1% increase in dynamic power. To the best of our knowledge, this is the first integrated mechanism to guarantee user-level interrupt delivery with a nanosecond-scale yet bounded worst-case latency. △ Less

Submitted 5 April, 2026; originally announced April 2026.

Persistent Cross-Attempt State Optimization for Repository-Level Code Generation

Authors: Ruwei Pan, Jiangshuai Wang, Qisheng Zhang, Yueheng Zhu, Linhao Wu, Zixiong Yang, Yakun Zhang, Lu Zhang, Hongyu Zhang

Abstract: Large language models (LLMs) have achieved substantial progress in repository-level code generation. However, solving the same repository-level task often requires multiple attempts, while existing methods still optimize each attempt in isolation and do not preserve or reuse task-specific state across attempts. In this paper, we propose LiveCoder, a novel framework for repository-level code genera… ▽ More Large language models (LLMs) have achieved substantial progress in repository-level code generation. However, solving the same repository-level task often requires multiple attempts, while existing methods still optimize each attempt in isolation and do not preserve or reuse task-specific state across attempts. In this paper, we propose LiveCoder, a novel framework for repository-level code generation based on cross-attempt knowledge optimization. LiveCoder maintains persistent task-specific state from prior attempts to guide subsequent generation. This state includes success knowledge, which captures reusable signals from previously strong repositories, failure knowledge, which records unsuccessful outcomes and their diagnostic signals, and a historical-best repository, which preserves the strongest result found so far and prevents regression. These components collectively transform repeated repository generation into a persistent, knowledge-driven optimization process. We evaluate LiveCoder using four frontier LLMs on two representative repository-level code generation benchmarks. Extensive experimental results demonstrate the effectiveness and efficiency of LiveCoder, improving the functional score by up to 22.94 percentage points, increasing repository reuse to 81.58%, and reducing cost by up to 53.63% on RAL-Bench while maintaining broadly stable non-functional quality. △ Less

Submitted 4 April, 2026; originally announced April 2026.

Toward Executable Repository-Level Code Generation via Environment Alignment

Authors: Ruwei Pan, Junlei Shen, Linhao Wu, Yueheng Zhu, Zixiong Yang, Yakun Zhang, Lu Zhang, Hongyu Zhang

Abstract: Large language models (LLMs) have achieved strong performance on code generation, but existing methods still struggle with repository-level code generation under executable validation. Under this evaluation setting, success is determined not by the plausibility of isolated code fragments, but by whether a generated multi-file repository can be successfully installed, have its dependencies and inte… ▽ More Large language models (LLMs) have achieved strong performance on code generation, but existing methods still struggle with repository-level code generation under executable validation. Under this evaluation setting, success is determined not by the plausibility of isolated code fragments, but by whether a generated multi-file repository can be successfully installed, have its dependencies and internal references resolved, be launched, and be validated in a real execution environment. To address this challenge, we propose EnvGraph, a framework for repository-level code generation that formulates repository executability as an environment alignment problem. EnvGraph jointly models two coupled conditions for successful repository execution, namely external dependency satisfaction and repository-internal reference resolution. It maintains a dual-layer environment representation, uses execution evidence to perform execution-evidence-based attribution, and guides repository generation through a unified targeted revision mechanism within an iterative alignment loop. We evaluate EnvGraph on repository-level code generation with three representative backbone LLMs and compare it against representative environment-aware and repository-level baselines. Experimental results show that EnvGraph consistently achieves the best performance on these repository-level benchmarks. In particular, it outperforms the strongest non-EnvGraph baseline by an absolute margin of 5.72--5.87 percentage points in Functional Correctness and 4.58--8.66 percentage points in Non-Functional Quality. △ Less

Submitted 4 April, 2026; originally announced April 2026.

Beyond Predefined Schemas: TRACE-KG for Context-Enriched Knowledge Graph Generation

Authors: Mohammad Sadeq Abolhasani, Yang Ba, Yixuan He, Rong Pan

Abstract: Knowledge graph generation typically relies either on predefined ontologies or on schema-free extraction. Ontology-driven pipelines enforce consistent typing but require costly schema design and maintenance, whereas schema-free methods often produce fragmented graphs with weak global organization, especially in long technical documents with dense, context-dependent information. We propose \textbf{… ▽ More Knowledge graph generation typically relies either on predefined ontologies or on schema-free extraction. Ontology-driven pipelines enforce consistent typing but require costly schema design and maintenance, whereas schema-free methods often produce fragmented graphs with weak global organization, especially in long technical documents with dense, context-dependent information. We propose \textbf{TRACE-KG} (\textbf{T}ext-d\textbf{R}iven schem\textbf{A} for \textbf{C}ontext-\textbf{E}nriched \textbf{K}nowledge \textbf{G}raphs), a framework that jointly constructs a context-enriched knowledge graph and an induced schema without assuming a predefined ontology. TRACE-KG captures conditional relations through structured qualifiers and organizes entities and relations using a data-driven schema that serves as a reusable semantic scaffold while preserving full traceability to the source evidence. Experiments show that TRACE-KG produces structurally coherent, traceable knowledge graphs and offers a practical alternative to both ontology-driven and schema-free construction pipelines. △ Less

Submitted 15 June, 2026; v1 submitted 3 April, 2026; originally announced April 2026.

Comments: Accepted at Graph Foundation Models at ICML 2026

Aligning Multimodal Sequential Recommendations via Robust Direct Preference Optimization with Sparse MoE

Authors: Hejin Huang, Jusheng Zhang, Kaitong Cai, Jian Wang, Rong Pan

Abstract: Preference-based alignment objectives have been widely adopted, from RLHF-style pairwise learning in large language models to emerging applications in recommender systems. Yet, existing work rarely examines how Direct Preference Optimization (DPO) behaves under implicit feedback, where unobserved items are not reliable negatives. We conduct systematic experiments on multimodal sequential recommend… ▽ More Preference-based alignment objectives have been widely adopted, from RLHF-style pairwise learning in large language models to emerging applications in recommender systems. Yet, existing work rarely examines how Direct Preference Optimization (DPO) behaves under implicit feedback, where unobserved items are not reliable negatives. We conduct systematic experiments on multimodal sequential recommendation to compare common negative-selection strategies and their interaction with DPO training. Our central finding is that a simple modification, replacing deterministic hard negatives with stochastic sampling from a dynamic top-K candidate pool, consistently improves ranking performance. We attribute its effectiveness to two factors: (1) reducing erroneous suppressive gradients caused by false negatives, and (2) retaining informative hard signals while smoothing optimization via controlled stochasticity. With an optional sparse Mixture-of-Experts encoder for efficient capacity scaling, RoDPO achieves up to 5.25% NDCG@5 on three Amazon benchmarks, with nearly unchanged inference cost. △ Less

Submitted 31 March, 2026; originally announced March 2026.

FedFG: Privacy-Preserving and Robust Federated Learning via Flow-Matching Generation

Authors: Ruiyang Wang, Rong Pan, Zhengan Yao

Abstract: Federated learning (FL) enables distributed clients to collaboratively train a global model using local private data. Nevertheless, recent studies show that conventional FL algorithms still exhibit deficiencies in privacy protection, and the server lacks a reliable and stable aggregation rule for updating the global model. This situation creates opportunities for adversaries: on the one hand, they… ▽ More Federated learning (FL) enables distributed clients to collaboratively train a global model using local private data. Nevertheless, recent studies show that conventional FL algorithms still exhibit deficiencies in privacy protection, and the server lacks a reliable and stable aggregation rule for updating the global model. This situation creates opportunities for adversaries: on the one hand, they may eavesdrop on uploaded gradients or model parameters, potentially leaking benign clients' private data; on the other hand, they may compromise clients to launch poisoning attacks that corrupt the global model. To balance accuracy and security, we propose FedFG, a robust FL framework based on flow-matching generation that simultaneously preserves client privacy and resists sophisticated poisoning attacks. On the client side, each local network is decoupled into a private feature extractor and a public classifier. Each client is further equipped with a flow-matching generator that replaces the extractor when interacting with the server, thereby protecting private features while learning an approximation of the underlying data distribution. Complementing the client-side design, the server employs a client-update verification scheme and a novel robust aggregation mechanism driven by synthetic samples produced by the flow-matching generator. Experiments on MNIST, FMNIST, and CIFAR-10 demonstrate that, compared with prior work, our approach adapts to multiple attack strategies and achieves higher accuracy while maintaining strong privacy protection. △ Less

Submitted 29 March, 2026; originally announced March 2026.

MemArchitect: A Policy Driven Memory Governance Layer

Authors: Lingavasan Suresh Kumar, Yang Ba, Rong Pan

Abstract: Persistent Large Language Model (LLM) agents expose a critical governance gap in memory management. Standard Retrieval-Augmented Generation (RAG) frameworks treat memory as passive storage, lacking mechanisms to resolve contradictions, enforce privacy, or prevent outdated information ("zombie memories") from contaminating the context window. We introduce MemArchitect, a governance layer that dec… ▽ More Persistent Large Language Model (LLM) agents expose a critical governance gap in memory management. Standard Retrieval-Augmented Generation (RAG) frameworks treat memory as passive storage, lacking mechanisms to resolve contradictions, enforce privacy, or prevent outdated information ("zombie memories") from contaminating the context window. We introduce MemArchitect, a governance layer that decouples memory lifecycle management from model weights. MemArchitect enforces explicit, rule-based policies, including memory decay, conflict resolution, and privacy controls. We demonstrate that governed memory consistently outperforms unmanaged memory in agentic settings, highlighting the necessity of structured memory governance for reliable and safe autonomous systems. △ Less

Submitted 18 March, 2026; originally announced March 2026.

Comments: This is an on going research work and will be updated periodically

HWE-Bench: Can Language Models Perform Board-level Schematic Designs?

Authors: Weibo Qiu, Yinhao Xiao, Runyu Pan

Abstract: Large Language Models (LLMs) have demonstrated significant potential in various engineering tasks, including software development, digital logic generation, and companion document maintenance. However, their ability to perform board-level circuit design is understudied, as this task requires a synergized understanding of real-world physics and Integrated Circuit (IC) datasheets, the latter compris… ▽ More Large Language Models (LLMs) have demonstrated significant potential in various engineering tasks, including software development, digital logic generation, and companion document maintenance. However, their ability to perform board-level circuit design is understudied, as this task requires a synergized understanding of real-world physics and Integrated Circuit (IC) datasheets, the latter comprising detailed specifications for individual components. To address this challenge, we propose \hweb, an evaluation framework that benchmarks the ability of LLMs to perform such designs. It consists of 300 board-level design tasks pulled from open-source and crowdsourcing platforms such as GitHub and OSHWLab, covering 8 application domains, and is complemented with a knowledge base of 2,914 real IC datasheets. For each task, the LLMs are tasked with generating a schematic from scratch, using the provided circuit functional requirements and a set of component datasheets as input. The resulting schematic will be checked against a static electrical rules, and then passed to a circuit simulator to verify its dynamic behavior. Our evaluation show that although current models achieve initial engineering usability and documentation understanding, they lack physical intuition, as the top-performing model achieved an overall pass rate of 8.15\%. We envision that advancements on \hweb\ will pave the way for the development of practical Electronic Design Automation (EDA) agents, revolutionizing the field of board-level design. △ Less

Submitted 18 March, 2026; originally announced March 2026.

Fusian: Multi-LoRA Fusion for Fine-Grained Continuous MBTI Personality Control in Large Language Models

Authors: Zehao Chen, Rong Pan

Abstract: Large Language Models (LLMs) have demonstrated impressive capabilities in simulating diverse human behaviors and personalities. However, existing methods for personality control, which include prompt engineering and standard Supervised Fine-Tuning (SFT), typically treat personality traits as discrete categories (e.g., "Extroverted" vs. "Introverted"), lacking the ability to precisely control the i… ▽ More Large Language Models (LLMs) have demonstrated impressive capabilities in simulating diverse human behaviors and personalities. However, existing methods for personality control, which include prompt engineering and standard Supervised Fine-Tuning (SFT), typically treat personality traits as discrete categories (e.g., "Extroverted" vs. "Introverted"), lacking the ability to precisely control the intensity of a trait on a continuous spectrum. In this paper, we introduce Fusian, a novel framework for fine-grained, continuous personality control in LLMs. Fusian operates in two stages: (1) Trajectory Collection, where we capture the dynamic evolution of personality adoption during SFT by saving a sequence of LoRA adapters, effectively mapping the continuous manifold of a trait; and (2) RL-based Dynamic Fusion, where we train a policy network using Reinforcement Learning to dynamically compute mixing weights for these frozen adapters. By sampling from a Dirichlet distribution parameterized by the policy network, Fusian fuses multiple adapters to align the model's output with a specific numerical target intensity. Experiments on the Qwen3-14B model demonstrate that Fusian achieves high precision in personality control, significantly outperforming baseline methods in aligning with user-specified trait intensities. △ Less

Submitted 16 March, 2026; originally announced March 2026.

A Comprehensive Evaluation of LLM Unlearning Robustness under Multi-Turn Interaction

Authors: Ruihao Pan, Suhang Wang

Abstract: Machine unlearning aims to remove the influence of specific training data from pre-trained models without retraining from scratch, and is increasingly important for large language models (LLMs) due to safety, privacy, and legal concerns. Although prior work primarily evaluates unlearning in static, single-turn settings, forgetting robustness under realistic interactive use remains underexplored. I… ▽ More Machine unlearning aims to remove the influence of specific training data from pre-trained models without retraining from scratch, and is increasingly important for large language models (LLMs) due to safety, privacy, and legal concerns. Although prior work primarily evaluates unlearning in static, single-turn settings, forgetting robustness under realistic interactive use remains underexplored. In this paper, we study whether unlearning remains stable in interactive environments by examining two common interaction patterns: self-correction and dialogue-conditioned querying. We find that knowledge appearing forgotten in static evaluation can often be recovered through interaction. Although stronger unlearning improves apparent robustness, it often results in behavioral rigidity rather than genuine knowledge erasure. Our findings suggest that static evaluation may overestimate real-world effectiveness and highlight the need for ensuring stable forgetting under interactive settings. △ Less

Submitted 14 March, 2026; v1 submitted 28 February, 2026; originally announced March 2026.

ClawMobile: Rethinking Smartphone-Native Agentic Systems

Authors: Hongchao Du, Shangyu Wu, Qiao Li, Riwei Pan, Jinheng Li, Youcheng Sun, Chun Jason Xue

Abstract: Smartphones represent a uniquely challenging environment for agentic systems. Unlike cloud or desktop settings, mobile devices combine constrained execution contexts, fragmented control interfaces, and rapidly changing application states. As large language models (LLMs) evolve from conversational assistants to action-oriented agents, achieving reliable smartphone-native autonomy requires rethinkin… ▽ More Smartphones represent a uniquely challenging environment for agentic systems. Unlike cloud or desktop settings, mobile devices combine constrained execution contexts, fragmented control interfaces, and rapidly changing application states. As large language models (LLMs) evolve from conversational assistants to action-oriented agents, achieving reliable smartphone-native autonomy requires rethinking how reasoning and control are composed. We introduce ClawMobile as a concrete exploration of this design space. ClawMobile adopts a hierarchical architecture that separates high-level language reasoning from structured, deterministic control pathways, improving execution stability and reproducibility on real devices. Using ClawMobile as a case study, we distill the design principles for mobile LLM runtimes and identify key challenges in efficiency, adaptability, and stability. We argue that building robust smartphone-native agentic systems demands principled coordination between probabilistic planning and deterministic system interfaces. The implementation is open-sourced~\footnote{https://github.com/ClawMobile/ClawMobile} to facilitate future exploration. △ Less

Submitted 11 April, 2026; v1 submitted 26 February, 2026; originally announced February 2026.

Comments: Accepted at EuroMLSys 2026, 7 pages, 1 figure

Measuring Dataset Diversity from a Geometric Perspective

Authors: Yang Ba, Mohammad Sadeq Abolhasani, Michelle V Mancenido, Rong Pan

Abstract: Diversity can be broadly defined as the presence of meaningful variation across elements, which can be viewed from multiple perspectives, including statistical variation and geometric structural richness in the dataset. Existing diversity metrics, such as feature-space dispersion and metric-space magnitude, primarily capture distributional variation or entropy, while largely neglecting the geometr… ▽ More Diversity can be broadly defined as the presence of meaningful variation across elements, which can be viewed from multiple perspectives, including statistical variation and geometric structural richness in the dataset. Existing diversity metrics, such as feature-space dispersion and metric-space magnitude, primarily capture distributional variation or entropy, while largely neglecting the geometric structure of datasets. To address this gap, we introduce a framework based on topological data analysis (TDA) and persistence landscapes (PLs) to extract and quantify geometric features from data. This approach provides a theoretically grounded means of measuring diversity beyond entropy, capturing the rich geometric and structural properties of datasets. Through extensive experiments across diverse modalities, we demonstrate that our proposed PLs-based diversity metric (PLDiv) is powerful, reliable, and interpretable, directly linking data diversity to its underlying geometry and offering a foundational tool for dataset construction, augmentation, and evaluation. △ Less

Submitted 9 February, 2026; originally announced February 2026.

Toward Functional and Non-Functional Evaluation of Application-Level Code Generation

Authors: Ruwei Pan, Yakun Zhang, Qingyuan Liang, Yueheng Zhu, Chao Liu, Lu Zhang, Hongyu Zhang

Abstract: Large language models (LLMs) have achieved strong performance on code generation. However, most prior evaluations focus on snippet-level outputs, such as function generation or repository completion. These settings do not fully evaluate application-level code generation, where the goal is to produce a runnable repository with coherent multi-file structure, dependency support, and end-to-end execut… ▽ More Large language models (LLMs) have achieved strong performance on code generation. However, most prior evaluations focus on snippet-level outputs, such as function generation or repository completion. These settings do not fully evaluate application-level code generation, where the goal is to produce a runnable repository with coherent multi-file structure, dependency support, and end-to-end executability. In addition, real-world software quality depends not only on functional correctness but also on non-functional quality attributes, such as maintainability and security. In this paper, we present RAL-Bench, a benchmark and evaluation framework for application-level code generation. For each task, RAL-Bench derives a concise natural-language requirement from a high-quality reference project, constructs black-box system tests for both functional correctness and non-functional quality attributes. It also retains only the candidate tests that pass on the reference repository. Under this unified evaluation protocol, functional correctness is measured by the system test pass rate, while non-functional quality is evaluated along five ISO/IEC 25010-inspired dimensions, with per-dimension diagnostics and reference-normalized scoring.We evaluate 16 frontier LLMs under a controlled zero-shot setting with greedy decoding. The results show that functional correctness remains the primary bottleneck in application-level code generation, while non-functional quality also remains challenging. Under our evaluation protocol, no model exceeds a 45\% functional score. These findings suggest that strong performance on existing code generation benchmarks does not yet translate to strong performance on application-level repository generation. This result highlights the need for evaluation settings that directly assess end-to-end repository generation rather than relying only on snippet-level success. △ Less

Submitted 30 March, 2026; v1 submitted 3 February, 2026; originally announced February 2026.

Usage, Effects and Requirements for AI Coding Assistants in the Enterprise: An Empirical Study

Authors: Maja Vukovic, Rangeet Pan, Tin Kam Ho, Rahul Krishna, Raju Pavuluri, Michele Merler

Abstract: The rise of large language models (LLMs) has accelerated the development of automated techniques and tools for supporting various software engineering tasks, e.g., program understanding, code generation, software testing, and program repair. As CodeLLMs are being employed toward automating these tasks, one question that arises, especially in enterprise settings, is whether these coding assistants… ▽ More The rise of large language models (LLMs) has accelerated the development of automated techniques and tools for supporting various software engineering tasks, e.g., program understanding, code generation, software testing, and program repair. As CodeLLMs are being employed toward automating these tasks, one question that arises, especially in enterprise settings, is whether these coding assistants and the code LLMs that power them are ready for real-world projects and enterprise use cases, and how do they impact the existing software engineering process and user experience. In this paper we survey 57 developers from different domains and with varying software engineering skill about their experience with AI coding assistants and CodeLLMs. We also reviewed 35 user surveys on the usage, experience and expectations of professionals and students using AI coding assistants and CodeLLMs. Based on our study findings and analysis of existing surveys, we discuss the requirements for AI-powered coding assistants. △ Less

Submitted 27 January, 2026; originally announced January 2026.

Comments: To appear in the 3rd International Workshop on Large Language Models For Code, co-located at ICSE, Rio de Janeiro, Brazil, 2026

LLM-powered Real-time Patent Citation Recommendation for Financial Technologies

Authors: Tianang Deng, Yu Deng, Tianchen Gao, Yonghong Hu, Rui Pan

Abstract: Rapid financial innovation has been accompanied by a sharp increase in patenting activity, making timely and comprehensive prior-art discovery more difficult. This problem is especially evident in financial technologies, where innovations develop quickly, patent collections grow continuously, and citation recommendation systems must be updated as new applications arrive. Existing patent retrieval… ▽ More Rapid financial innovation has been accompanied by a sharp increase in patenting activity, making timely and comprehensive prior-art discovery more difficult. This problem is especially evident in financial technologies, where innovations develop quickly, patent collections grow continuously, and citation recommendation systems must be updated as new applications arrive. Existing patent retrieval and citation recommendation methods typically rely on static indexes or periodic retraining, which limits their ability to operate effectively in such dynamic settings. In this study, we propose a real-time patent citation recommendation framework designed for large and fast-changing financial patent corpora. Using a dataset of 428,843 financial patents granted by the China National Intellectual Property Administration (CNIPA) between 2000 and 2024, we build a three-stage recommendation pipeline. The pipeline uses large language model (LLM) embeddings to represent the semantic content of patent abstracts, applies efficient approximate nearest-neighbor search to construct a manageable candidate set, and ranks candidates by semantic similarity to produce top-k citation recommendations. In addition to improving recommendation accuracy, the proposed framework directly addresses the dynamic nature of patent systems. By using an incremental indexing strategy based on hierarchical navigable small-world (HNSW) graphs, newly issued patents can be added without rebuilding the entire index. A rolling day-by-day update experiment shows that incremental updating improves recall while substantially reducing computational cost compared with rebuild-based indexing. The proposed method also consistently outperforms traditional text-based baselines and alternative nearest-neighbor retrieval approaches. △ Less

Submitted 23 January, 2026; originally announced January 2026.

PhysProver: Advancing Automatic Theorem Proving for Physics

Authors: Hanning Zhang, Ruida Wang, Rui Pan, Wenyuan Wang, Bingxu Meng, Tong Zhang

Abstract: The combination of verifiable languages and LLMs has significantly influenced both the mathematical and computer science communities because it provides a rigorous foundation for theorem proving. Recent advancements in the field provide foundation models and sophisticated agentic systems pushing the boundaries of formal mathematical reasoning to approach the natural language capability of LLMs. Ho… ▽ More The combination of verifiable languages and LLMs has significantly influenced both the mathematical and computer science communities because it provides a rigorous foundation for theorem proving. Recent advancements in the field provide foundation models and sophisticated agentic systems pushing the boundaries of formal mathematical reasoning to approach the natural language capability of LLMs. However, little attention has been given to the formal physics reasoning, which also heavily relies on similar problem-solving and theorem-proving frameworks. To solve this problem, this paper presents, to the best of our knowledge, the first approach to enhance formal theorem proving in the physics domain. We compose a dedicated dataset PhysLeanData for the task. It is composed of theorems sampled from PhysLean and data generated by a conjecture-based formal data generation pipeline. In the training pipeline, we leverage DeepSeek-Prover-V2-7B, a strong open-source mathematical theorem prover, and apply Reinforcement Learning with Verifiable Rewards (RLVR) to train our model PhysProver. Comprehensive experiments demonstrate that, using only $\sim$5K training samples, PhysProver achieves an overall 2.4\% improvement in multiple sub-domains. Furthermore, after formal physics training, we observe 1.3\% gains on the MiniF2F-Test benchmark, which indicates non-trivial generalization beyond physics domains and enhancement for formal math capability as well. The results highlight the effectiveness and efficiency of our approach, which provides a paradigm for extending formal provers outside mathematical domains. To foster further research, we will release both our dataset and model to the community. △ Less

Submitted 22 January, 2026; originally announced January 2026.

Comments: Preprint

ArchAgent: Scalable Legacy Software Architecture Recovery with LLMs

Authors: Rusheng Pan, Bingcheng Mao, Tianyi Ma, Zhenhua Ling

Abstract: Recovering accurate architecture from large-scale legacy software is hindered by architectural drift, missing relations, and the limited context of Large Language Models (LLMs). We present ArchAgent, a scalable agent-based framework that combines static analysis, adaptive code segmentation, and LLM-powered synthesis to reconstruct multiview, business-aligned architectures from cross-repository cod… ▽ More Recovering accurate architecture from large-scale legacy software is hindered by architectural drift, missing relations, and the limited context of Large Language Models (LLMs). We present ArchAgent, a scalable agent-based framework that combines static analysis, adaptive code segmentation, and LLM-powered synthesis to reconstruct multiview, business-aligned architectures from cross-repository codebases. ArchAgent introduces scalable diagram generation with contextual pruning and integrates cross-repository data to identify business-critical modules. Evaluations of typical large-scale GitHub projects show significant improvements over existing benchmarks. An ablation study confirms that dependency context improves the accuracy of generated architectures of production-level repositories, and a real-world case study demonstrates effective recovery of critical business logics from legacy projects. The dataset is available at https://github.com/panrusheng/arch-eval-benchmark. △ Less

Submitted 19 January, 2026; originally announced January 2026.

Comments: to be published in ICASSP 2026

Investigating Tool-Memory Conflicts in Tool-Augmented LLMs

Authors: Jiali Cheng, Rui Pan, Hadi Amiri

Abstract: Tool-augmented large language models (LLMs) have powered many applications. However, they are likely to suffer from knowledge conflict. In this paper, we propose a new type of knowledge conflict -- Tool-Memory Conflict (TMC), where the internal parametric knowledge contradicts with the external tool knowledge for tool-augmented LLMs. We find that existing LLMs, though powerful, suffer from TMC, es… ▽ More Tool-augmented large language models (LLMs) have powered many applications. However, they are likely to suffer from knowledge conflict. In this paper, we propose a new type of knowledge conflict -- Tool-Memory Conflict (TMC), where the internal parametric knowledge contradicts with the external tool knowledge for tool-augmented LLMs. We find that existing LLMs, though powerful, suffer from TMC, especially on STEM-related tasks. We also uncover that under different conditions, tool knowledge and parametric knowledge may be prioritized differently. We then evaluate existing conflict resolving techniques, including prompting-based and RAG-based methods. Results show that none of these approaches can effectively resolve tool-memory conflicts. △ Less

Submitted 13 January, 2026; originally announced January 2026.

Comments: R2-FM Workshop @ ICML 2025

Fail Fast, Win Big: Rethinking the Drafting Strategy in Speculative Decoding via Diffusion LLMs

Authors: Rui Pan, Zhuofu Chen, Hongyi Liu, Arvind Krishnamurthy, Ravi Netravali

Abstract: Diffusion Large Language Models (dLLMs) offer fast, parallel token generation, but their standalone use is plagued by an inherent efficiency-quality tradeoff. We show that, if carefully applied, the attributes of dLLMs can actually be a strength for drafters in speculative decoding with autoregressive (AR) verifiers. Our core insight is that dLLM's speed from parallel decoding drastically lowers t… ▽ More Diffusion Large Language Models (dLLMs) offer fast, parallel token generation, but their standalone use is plagued by an inherent efficiency-quality tradeoff. We show that, if carefully applied, the attributes of dLLMs can actually be a strength for drafters in speculative decoding with autoregressive (AR) verifiers. Our core insight is that dLLM's speed from parallel decoding drastically lowers the risk of costly rejections, providing a practical mechanism to effectively realize the (elusive) lengthy drafts that lead to large speedups with speculative decoding. We present FailFast, a dLLM-based speculative decoding framework that realizes this approach by dynamically adapting its speculation length. It "fails fast" by spending minimal compute in hard-to-speculate regions to shrink speculation latency and "wins big" by aggressively extending draft lengths in easier regions to reduce verification latency (in many cases, speculating and accepting 70 tokens at a time!). Without any fine-tuning, FailFast delivers lossless acceleration of AR LLMs and achieves up to 4.9$\times$ speedup over vanilla decoding, 1.7$\times$ over the best naive dLLM drafter, and 1.7$\times$ over EAGLE-3 across diverse models and workloads. We open-source FailFast at https://github.com/ruipeterpan/failfast. △ Less

Submitted 28 January, 2026; v1 submitted 23 December, 2025; originally announced December 2025.

DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models

Authors: DeepSeek-AI, Aixin Liu, Aoxue Mei, Bangcai Lin, Bing Xue, Bingxuan Wang, Bingzheng Xu, Bochao Wu, Bowei Zhang, Chaofan Lin, Chen Dong, Chengda Lu, Chenggang Zhao, Chengqi Deng, Chenhao Xu, Chong Ruan, Damai Dai, Daya Guo, Dejian Yang, Deli Chen, Erhang Li, Fangqi Zhou, Fangyun Lin, Fucong Dai, Guangbo Hao , et al. (239 additional authors not shown)

Abstract: We introduce DeepSeek-V3.2, a model that harmonizes high computational efficiency with superior reasoning and agent performance. The key technical breakthroughs of DeepSeek-V3.2 are as follows: (1) DeepSeek Sparse Attention (DSA): We introduce DSA, an efficient attention mechanism that substantially reduces computational complexity while preserving model performance in long-context scenarios. (2)… ▽ More We introduce DeepSeek-V3.2, a model that harmonizes high computational efficiency with superior reasoning and agent performance. The key technical breakthroughs of DeepSeek-V3.2 are as follows: (1) DeepSeek Sparse Attention (DSA): We introduce DSA, an efficient attention mechanism that substantially reduces computational complexity while preserving model performance in long-context scenarios. (2) Scalable Reinforcement Learning Framework: By implementing a robust reinforcement learning protocol and scaling post-training compute, DeepSeek-V3.2 performs comparably to GPT-5. Notably, our high-compute variant, DeepSeek-V3.2-Speciale, surpasses GPT-5 and exhibits reasoning proficiency on par with Gemini-3.0-Pro, achieving gold-medal performance in both the 2025 International Mathematical Olympiad (IMO) and the International Olympiad in Informatics (IOI). (3) Large-Scale Agentic Task Synthesis Pipeline: To integrate reasoning into tool-use scenarios, we developed a novel synthesis pipeline that systematically generates training data at scale. This methodology facilitates scalable agentic post-training, yielding substantial improvements in generalization and instruction-following robustness within complex, interactive environments. △ Less

Submitted 2 December, 2025; originally announced December 2025.

Advancing Automated In-Isolation Validation in Repository-Level Code Translation

Authors: Kaiyao Ke, Ali Reza Ibrahimzada, Rangeet Pan, Saurabh Sinha, Reyhaneh Jabbarvand

Abstract: Repository-level code translation aims to migrate entire repositories across programming languages while preserving functionality automatically. Despite advancements in repository-level code translation, validating the translations remains challenging. This paper proposes TRAM, which combines context-aware type resolution with mock-based in-isolation validation to achieve high-quality translations… ▽ More Repository-level code translation aims to migrate entire repositories across programming languages while preserving functionality automatically. Despite advancements in repository-level code translation, validating the translations remains challenging. This paper proposes TRAM, which combines context-aware type resolution with mock-based in-isolation validation to achieve high-quality translations between programming languages. Prior to translation, TRAM retrieves API documentation and contextual code information for each variable type in the source language. It then prompts a large language model (LLM) with retrieved contextual information to resolve type mappings across languages with precise semantic interpretations. Using the automatically constructed type mapping, TRAM employs a custom serialization/deserialization workflow that automatically constructs equivalent mock objects in the target language. This enables each method fragment to be validated in isolation, without the high cost of using agents for translation validation, or the heavy manual effort required by existing approaches that rely on language interoperability. TRAM demonstrates state-of-the-art performance in Java-to-Python translation, underscoring the effectiveness of its integration of RAG-based type resolution with reliable in-isolation validation. △ Less

Submitted 23 December, 2025; v1 submitted 26 November, 2025; originally announced November 2025.

Denoising Refinement Diffusion Models for Simultaneous Generation of Multi-scale Mobile Network Traffic

Authors: Xiaoqian Qi, Haoye Chai, Sichang Liu, Lei Yue, Raoyuan Pan, Yue Wang, Yong Li

Abstract: The planning, management, and resource scheduling of cellular mobile networks require joint estimation of mobile traffic across different layers and nodes. Mobile traffic generation can proactively anticipate user demands and capture the dynamics of network load. However, existing methods mainly focus on generating traffic at a single spatiotemporal resolution, making it difficult to jointly model… ▽ More The planning, management, and resource scheduling of cellular mobile networks require joint estimation of mobile traffic across different layers and nodes. Mobile traffic generation can proactively anticipate user demands and capture the dynamics of network load. However, existing methods mainly focus on generating traffic at a single spatiotemporal resolution, making it difficult to jointly model multi-scale traffic patterns. In this paper, we propose ZoomDiff, a diffusion-based model for multi-scale mobile traffic generation. ZoomDiff maps urban environmental context into mobile traffic with multiple spatial and temporal resolutions through a set of customized Denoising Refinement Diffusion Models (DRDM). DRDM employs a multi-stage noise-adding and denoising mechanism, enabling different stages to generate traffic at distinct spatiotemporal resolutions. This design aligns the progressive denoising process with hierarchical network layers, including base stations, cells, and grids of varying granularities. Experiments on real-world mobile traffic datasets show that ZoomDiff achieves at least an 18.4% improvement over state-of-the-art baselines in multi-scale traffic generation tasks. Moreover, ZoomDiff demonstrates strong efficiency and cross-city generalization, highlighting its potential as a powerful generative framework for modeling multi-scale mobile network dynamics. △ Less

Submitted 24 November, 2025; v1 submitted 29 October, 2025; originally announced November 2025.

SAINT: Service-level Integration Test Generation with Program Analysis and LLM-based Agents

Authors: Rangeet Pan, Raju Pavuluri, Ruikai Huang, Rahul Krishna, Tyler Stennett, Alessandro Orso, Saurabh SInha

Abstract: Enterprise applications are typically tested at multiple levels, with service-level testing playing an important role in validating application functionality. Existing service-level testing tools, especially for RESTful APIs, often employ fuzzing and/or depend on OpenAPI specifications which are not readily available in real-world enterprise codebases. Moreover, these tools are limited in their ab… ▽ More Enterprise applications are typically tested at multiple levels, with service-level testing playing an important role in validating application functionality. Existing service-level testing tools, especially for RESTful APIs, often employ fuzzing and/or depend on OpenAPI specifications which are not readily available in real-world enterprise codebases. Moreover, these tools are limited in their ability to generate functional tests that effectively exercise meaningful scenarios. In this work, we present SAINT, a novel white-box testing approach for service-level testing of enterprise Java applications. SAINT combines static analysis, large language models (LLMs), and LLM-based agents to automatically generate endpoint and scenario-based tests. The approach builds two key models: an endpoint model, capturing syntactic and semantic information about service endpoints, and an operation dependency graph, capturing inter-endpoint ordering constraints. SAINT then employs LLM-based agents to generate tests. Endpoint-focused tests aim to maximize code and database interaction coverage. Scenario-based tests are synthesized by extracting application use cases from code and refining them into executable tests via planning, action, and reflection phases of the agentic loop. We evaluated SAINT on eight Java applications, including a proprietary enterprise application. Our results illustrate the effectiveness of SAINT in coverage, fault detection, and scenario generation. Moreover, a developer survey provides strong endorsement of the scenario-based tests generated by SAINT. Overall, our work shows that combining static analysis with agentic LLM workflows enables more effective, functional, and developer-aligned service-level test generation. △ Less

Submitted 24 May, 2026; v1 submitted 17 November, 2025; originally announced November 2025.

Comments: Accepted at ICSE'26

OSGym: Scalable OS Infra for Computer Use Agents

Authors: Zengyi Qin, Jinyuan Chen, Yunze Man, Shengcao Cao, Ziqi Pang, Zhuoyuan Wang, Han Fang, Ling Zhu, Zixin Xie, Zibu Wei, Tianshu Ran, Haoran Geng, Ray Pan, Qizhen Sun, Zachary Bright, Yuyang Cai, Chongye Yang, Jiace Zhao, Tianrui Liu, Han Cao, Yeyang Zhou, Rui Wang, Song Wang, Xiang Ren, Bo Zhang , et al. (3 additional authors not shown)

Abstract: Training computer use agents requires full-featured OS sandboxes with GUI environments, which consume substantial hardware resources as the number of sandboxes scales. Stochastic errors arising from diverse software execution within these sandboxes further demand robust infrastructure design and reliable error recovery. We present OSGym, a scalable OS environment infrastructure for computer use ag… ▽ More Training computer use agents requires full-featured OS sandboxes with GUI environments, which consume substantial hardware resources as the number of sandboxes scales. Stochastic errors arising from diverse software execution within these sandboxes further demand robust infrastructure design and reliable error recovery. We present OSGym, a scalable OS environment infrastructure for computer use agents, built around these key optimization strategies: (1) Decentralized OS state management, which isolates failures to individual replicas and significantly enhances overall system reliability; (2) Hardware-aware OS replica orchestration, which addresses CPU-bounded scaling bottlenecks and substantially reduces compute overhead; (3) KVM virtualization with copy-on-write disk management, which shares a common bootable disk across VM instances and provisions only instance-specific modifications, reducing physical disk consumption by 88% and increasing disk provisioning speed by 37 times; and (4) Robust container pool with multi-layer fault recovery. Together, these optimizations yield strong scalability and resource efficiency: OSGym manages over a thousand OS replicas under constrained resources, supports parallel trajectory generation at 1420 multi-turn trajectories per minute, and reduces per-replica cost to 0.2-0.3 USD per day, a 90% reduction over standard deployment. Our experiments validate OSGym across end-to-end pipelines for data collection and training for computer use agents. We believe OSGym establishes a new foundation for scalable, general-purpose computer use agent research. △ Less

Submitted 1 April, 2026; v1 submitted 11 November, 2025; originally announced November 2025.

NegoCollab: A Common Representation Negotiation Approach for Heterogeneous Collaborative Perception

Authors: Congzhang Shao, Quan Yuan, Guiyang Luo, Yue Hu, Danni Wang, Yilin Liu, Rui Pan, Bo Chen, Jinglin Li

Abstract: Collaborative perception improves task performance by expanding the perception range through information sharing among agents. . Immutable heterogeneity poses a significant challenge in collaborative perception, as participating agents may employ different and fixed perception models. This leads to domain gaps in the intermediate features shared among agents, consequently degrading collaborative p… ▽ More Collaborative perception improves task performance by expanding the perception range through information sharing among agents. . Immutable heterogeneity poses a significant challenge in collaborative perception, as participating agents may employ different and fixed perception models. This leads to domain gaps in the intermediate features shared among agents, consequently degrading collaborative performance. Aligning the features of all agents to a common representation can eliminate domain gaps with low training cost. However, in existing methods, the common representation is designated as the representation of a specific agent, making it difficult for agents with significant domain discrepancies from this specific agent to achieve proper alignment. This paper proposes NegoCollab, a heterogeneous collaboration method based on the negotiated common representation. It introduces a negotiator during training to derive the common representation from the local representations of each modality's agent, effectively reducing the inherent domain gap with the various local representations. In NegoCollab, the mutual transformation of features between the local representation space and the common representation space is achieved by a pair of sender and receiver. To better align local representations to the common representation containing multimodal information, we introduce structural alignment loss and pragmatic alignment loss in addition to the distribution alignment loss to supervise the training. This enables the knowledge in the common representation to be fully distilled into the sender. △ Less

Submitted 30 January, 2026; v1 submitted 31 October, 2025; originally announced October 2025.

Comments: 23 pages, Accepted by NeurIPS 2025