Artificial Intelligence · Cross-listed from cs.CR
RustMizan: A Compilable, Contamination-Aware Benchmarking Framework for Rust Vulnerabilities
Tarek Elsayed, Shiping Yang, Eunsong Koh et al. (+13 more)
LLM
agents are increasingly applied to vulnerability analysis, but existing
benchmarks have not kept pace. They typically rely on small non-compilable snippets, focus on binary classification (vulnerable or not), and do not account for the risk that publicly-released
datasets are part of model training corpora. We introduce RustMizan, a benchmarking framework for Rust vulnerability analysis that addresses these gaps. RustMizan contains compilable code variants at the crate, file, and function levels, with annotations for binary vulnerability detection, CWE classification, and function- and line-level localization. A paired mutation framework produces semantics-preserving code mutants for contamination testing and robustness probing. Across four frontier models in an agentic setup with command-line access, binary classification sits in the 56-65% range, but line localization F1 stays near 20%, and
adversarial cues drop line F1 by about 27%.
Read the full abstract ▾
All authors Tarek Elsayed, Shiping Yang, Eunsong Koh, Sanika Goyal, Vincent Huang, Paul Ngo, Nathan Young, Mohammad Omidvar Tehrani, Alvyn Kang, Arnell Kang, Zeyu Chen, Angélica Moreira, Xuan Feng, Angel X. Chang, Nick Sumner, Steven Y. Ko
Artificial Intelligence · Cross-listed from cs.CL
Turning Off-Policy Tokens On-Policy: A Plug-in Approach for Improving LLM Alignment
Yu Li, Xiuyu Li, Mingyang Yi et al. (+4 more)
Reinforcement learning (RL) post-training for large
language models (LLMs) follows a efficient paradigm of "rollout then update", which inevitably results in off-policy training data. To resolve this, Importance sampling (IS) is proposed, while the
token-level ratios compound over long sequences, causing severe variance exploded. A natural idea is "transferring" these off-policy
token into on-policy
token, so that the importance scores for correction are unnecessary. Following this idea, we propose Selective Importance Sampling (SIS), which is inspired by rejection sampling. Concretely, SIS implements by viewing off-policy model as proposal distribution, and implement a
token-level rejection test: accepted
tokens are viewed as on-policy, so that receive unit importance score, while rejected
tokens retain the standard IS correction. Our proposed SIS is theoretically proved reducing the gap between
token-level and sequence-level off-policy gradient
estimators. The SIS acts as a plug-in that only modifies the importance ratio in the policy loss, adding negligible wall-clock overhead, and can be combine with a vast vary of RL post-training
algorithms. Experiments on dense and MoE LLMs across math and
agent benchmarks show that SIS consistently improves all objectives, while providing substantially stronger robustness under off-policy data.
Read the full abstract ▾
All authors Yu Li, Xiuyu Li, Mingyang Yi, Jiaxing Wang, Liangxu Zhang, Zhaolong Xing, Zhen Chen
Artificial Intelligence · Cross-listed from cs.SE
Dashboard2Code: Evaluating Multimodal Models on Reconstructing Interactive Dashboards
Tianhao Niu, Ziyu Han, Qiguang Chen et al. (+5 more)
Automatic data visualization generation has advanced rapidly with multi-modal large
language models, yet existing efforts largely focus on static charts and overlook the interactive dashboards commonly used for real-world data exploration. We introduce Dashboard2Code, a novel task that requires a model to proactively explore an interactive dashboard, acquire and integrate feedback from its own interactions (e.g., clicking and filtering), and generate code that reproduces the target dashboard. To support comprehensive evaluation, we present DashboardMimic, the first Plotly+Dash
benchmark for Dashboard2Code, comprising 180 carefully designed and manually verified dashboard-code pairs spanning three difficulty levels and covering eight common real-world interaction patterns. We further propose an automated evaluation framework tailored to dashboards that combines code semantic analysis with dynamic interaction-based testing to assess visual and interaction consistency, showing strong agreement with human judgments. Experiments across a range of open- and closed-source multi-modal models reveal that even the strongest systems struggle on high-complexity dashboards and that a substantial performance gap remains between open-source and closed-source models on the Dashboard2Code task.
Read the full abstract ▾
All authors Tianhao Niu, Ziyu Han, Qiguang Chen, Shiqi Zhou, Baocai Shan, Hengjie Fang, Qingfu Zhu, Wanxiang Che
Artificial Intelligence
FORGE: Research-Trajectory Hijacking Attacks on Deep Research Agents
Yue Pan, Ziheng Zhang, Junxiang Lei et al. (+3 more)
Deep research
agents decompose open-ended queries into subtasks, retrieve web evidence over multiple rounds, and synthesize long-form reports. This workflow creates a planning-layer poisoning surface:
adversarial documents that enter the
retrieval pool can steer follow-up questions and turn a local injection into report-level contamination. We present FORGE (Fabricated Orchestrated
Reasoning chain for
aGent Exploitation), a two-level attack that combines intra-document
reasoning fabrication with inter-document chain coordination to hijack subtask planning. We further introduce the PRISM metric, which weights infected report claims by cognitive type, and Root Query Anchoring, a lightweight defense that ties recursive follow-up generation to the root query. Across 25 queries, Network FORGE reaches 26.4% PRISM with five injected documents and exhibits depth migration, in which recursive synthesis shifts poisoned content from overt framing into factual premises. On the 10-query defense subset, RQA (Root Query Anchoring) reduces PRISM from 38.5% to 18.3%.
Read the full abstract ▾
All authors Yue Pan, Ziheng Zhang, Junxiang Lei, Changhao Jia, Qingyi Si, Hongcheng Guo
Artificial Intelligence · Cross-listed from cs.RO
Geometry-Aware Motion Latents for Learning Robust Manipulation Policies
Yunchao Zhang, Yijia Weng, Ruizhe Liu et al. (+3 more)
Learning motion latents for robotic manipulation heavily relies on extracting motion patterns from visual sequences, yet effective action abstractions require understanding three-dimensional
geometric transformations. Here, we introduce GeoMoLa (Geometry-Aware Motion Latents), which learns discrete motion latent codes by predicting how point clouds evolve during manipulation rather than reconstructing visual observations. This four-dimensional objective – spatial geometry changing through time – forces latent representations to encode actual physical motion rather than appearance patterns. GeoMoLa achieves state-of-the-art performance using only single-view RGB-D input, while existing methods require multi-view reconstruction, succeeding across diverse manipulation
benchmarks. Our ablations reveal that
geometric prediction is the key to driving performance, quantitatively validating that manipulation depends on spatial understanding. Furthermore, the learned codes exhibit effective motion abstraction: applying them to novel scenes produces physically consistent transformations regardless of visual context. Our real-world experiments also confirm this robustness capability, achieving robust manipulation with minimal demonstrations in cluttered environments where
geometric reasoning determines success. Thus, we demonstrate that effective motion latents for
robot control can better emerge from understanding motion through its three-dimensional effects rather than pixel-level patterns.
Read the full abstract ▾
All authors Yunchao Zhang, Yijia Weng, Ruizhe Liu, Ming Hu, Leonidas Guibas, Yanchao Yang
Artificial Intelligence · Cross-listed from cs.LG
RSPO: Reward-Swap Policy Optimization for Multi-Turn LLM Agents
Qiang Liu, Taian Guo, Ruizhi Qiao et al. (+1 more)
Reinforcement learning holds significant potential for training large
language models (LLMs) to handle multi-turn interactive tasks. However, in long-horizon, multi-turn tasks characterized by sparse outcome rewards, directly training with outcome rewards often results in slow convergence due to the sparsity of signals and the lack of fine-grained feedback. Furthermore, the model may fail to learn successful trajectories that are not sampled during training, thereby limiting its performance. Conversely, while employing customized dense process rewards provides richer signals and accelerates convergence, these surrogate rewards may exhibit potential misalignment with the ground-truth outcome rewards. This inconsistency can bias the training direction and ultimately degrade the model's final performance. In this work, we propose Reward-Swap Policy Optimization (RSPO), a method designed to leverage the rich information from dense process rewards to facilitate training with outcome rewards. By utilizing a reward-swap mechanism, RSPO ensures the diversity of sampled trajectories while guaranteeing consistency between the optimization objective and the true outcome rewards, thereby elevating the performance ceiling of the model. We conduct extensive experiments on two challenging
agent benchmarks, WebShop and ALFWorld. By applying our method to various reinforcement learning
algorithms, including GRPO, PPO, and GiGPO, we demonstrate that RSPO achieves consistent performance improvements across different baselines and
benchmarks.
Read the full abstract ▾
All authors Qiang Liu, Taian Guo, Ruizhi Qiao, Xing Sun
Artificial Intelligence
Integrated Altruistic and Fairness Preference Induces Advanced Mutual Cooperation in Sequential Social Dilemmas
Yu Wei, Yukiko Ogura, Yoshiyuki Ohmura et al. (+3 more)
Inducing cooperation among distributed
agents is still a difficult problem in the field of multi-
agent reinforcement learning (MARL), particularly in social dilemma situations. There, individual interests are misaligned with the common good and individual rationality leads to suboptimal group outcomes. In contrast, humans are able to achieve cooperation with one another in such situations. A common explanation for such cooperative behavior is that individuals have social preferences. In order to achieve cooperation in MARL, we design a new utility function integrating altruistic preferences (incentive for other's reward) and
fairness preferences (incentive for equality) from social psychology and behavioral economics, namely, Altruistic and
Fairness Preference (AFP), a reward-sharing mechanism which converts one's own and other's rewards to incentives for cooperative behavior. We performed comparative experiments with standard RL and inequity aversion
agents in two challenging sequential social dilemma games, and showed that AFP
agents successfully achieved mutual cooperation with more collective rewards and higher equity than the baselines. To further understand the progression of AFP during training, we subsequently explore the effects of altruistic preferences and
fairness preferences on
agents' behavior. The results suggest that altruistic preferences encourage
agents to contribute to the public goods, and
fairness preferences induce mutual behavior between
agents.
Read the full abstract ▾
All authors Yu Wei, Yukiko Ogura, Yoshiyuki Ohmura, Ildefons Magrans de Abril, Hoshinori Kanazawa, Yasuo Kuniyoshi
Artificial Intelligence · Cross-listed from econ.TH
Strategic Buying Agents
Mingyang Fu and Ming Hu
Agentic AI is shifting online shopping from search toward delegated purchasing, where autonomous buying
agents monitor
markets and decide when to buy on a consumer's behalf. We study the design of such strategic buying
agents, which must decide when to purchase within a finite shopping window, translating price observations, the remaining time horizon, and beliefs about future price changes into a purchase policy. We formulate this problem across three information regimes: stationary,
Bayesian, and robust, and treat the resulting optimal policies as a policy menu for implementation. In the stationary regime, price adjustments follow a Poisson arrival process with a known post-adjustment price distribution; the optimal policy is a dynamic purchase-threshold rule, with the threshold governed by an ordinary differential equation. In the
Bayesian regime, the adjustment intensity is known, but the price-adjustment distribution is uncertain; the optimal rule remains threshold-based, now depending on posterior beliefs, and we bound the value of knowing the true distribution. In the robust regime, the
agent has only price bounds and seeks worst-case protection; randomized threshold policies achieve optimal competitive-ratio and minimax-regret guarantees. We evaluate the proposed policies on Amazon price histories from Keepa (367 items, 48,933 timestamped observations) and examine their integration into
language-model buying
agents. The stationary and
Bayesian policies perform competitively on mean normalized consumer surplus despite their stylized assumptions, while the robust policy performs best at the distribution's 10th percentile. Results suggest
language models are better suited to selecting among regimes and calibration samples than to making buy-or-wait decisions directly.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.SE
AI Agent Pull Requests on GitHub: Frequency, Structure, and Merge Conflict Rates
George Xu, Arjun Subramanian and Nithilan Karthik
AI coding
agents may generate and submit Pull Requests (PRs) to the same repository at the same time. However, research concerning the extent of concurrent submission by AI coding
agents to a common repository does not exist. This paper uses the AIDev-pop
dataset (33,596 PRs in 2,807 repositories) to provide the first empirical examination of the prevalence of concurrent submission using PRs authored by
agents. We report that when considering exact temporal overlap, 40.2% of repositories contain co-active
agent-authored PR pairs; further, the co-active pairs account for 79.4% of all PRs generated by an AI
agent. When we examine co-activity within a one week collaboration window, the percentages are increased to 53.4% and 95.0%, respectively. For the majority of the co-active PR pairs (underlying the vast majority of which are intra-
agent authored), both PRs were authored by the same
agent, while only 0.5% of co-active pairs were cross-
agent, and occurred in only 122 out of 2807 total repositories examined (or approximately 4.3%). Additionally, we replayed actual three way git merges on 747 unique co-active pairs (one per repository), and computed the percentage of textual conflict encountered during the merge operation to combine the two PRs in each pair. We observed that the percentage of textual conflict encountered was significantly higher for cross-
agent pairs compared to intra-
agent pairs: 41.7% vs. 19.8%, respectively, with non-overlapping 95% confidence intervals. Lastly, we developed a classification system based on the detection of conflict reported by git, and determined that the majority of conflicts resulted from modifications to source code files (84.4% of conflicted files) and not dependency manifest files; further, nearly 42% of conflicts we observed were structural (i.e., modify/delete or add/add).
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.LG
URSA: Chemistry-Aware Benchmark for Utilitarian Retrosynthesis Assessment
Bogdan Zagribelnyy, Ivan Ilin, Nikita Bondarev et al. (+7 more)
Synthesis planning aiming to find pathways of reactions for a target molecule is one of the most important and challenging tasks in drug discovery. Recent progress has produced both specialized deep-learning retrosynthesis systems and general-purpose large
language models, but objective comparison remains difficult due to the lack of flexible, chemically interpretable benchmarking protocols. In the current study, we are introducing the URSA (Utilitarian RetroSynthesis Assessment) evaluation framework that provides the opportunity to
benchmark the synthetic routes not only from a formal perspective, such as convergence to commercially available starting materials, but also from a chemical plausibility perspective, mimicking the way expert chemists evaluate the reactions and routes. The study covers a comprehensive evaluation of both conventional end-to-end retrosynthesis solutions and LLMs for the synthesis planning task on a set of novel, diverse target molecules with undisclosed synthetic routes, which represent realistic tasks in the daily drug design routine. We find that while LLMs can support high-level strategic planning, they currently underperform specialized retrosynthesis models in reliably solving synthesis planning tasks.
Read the full abstract ▾
All authors Bogdan Zagribelnyy, Ivan Ilin, Nikita Bondarev, Anton Morgunov, Arkadii Lin, Maksim Kuznetsov, Rim Shayakhmetov, Vladimir Aladinskiy, Alex Aliper, Alex Zhavoronkov
Artificial Intelligence · Cross-listed from cs.CL
ToolFailBench: Diagnosing Tool-Use Failures in LLM Agents
Harsh Soni
Tool calling is central to modern
language model
agents, but aggregate
benchmark scores often hide where tool use fails. A model that never calls a needed tool and a model that calls the tool but ignores the result can look similar under final task accuracy. We introduce ToolFailBench, a diagnostic
benchmark for measuring tool-use failures across 1,000 tasks in finance, medicine, law, cybersecurity, and real estate. Tool-required tasks return values the model wouldn't guess, forcing it to trust the tool while control tasks attach the same tools but should be answered directly. We label each trace with Tool-Skip, Result-Ignore, Output-Fabrication, and Unnecessary-Tool-Use, using a rule classifier and two LLM judges aggregated by majority vote. Across 19 headline models, the best reaches 86.33% Clean Tool-Use Rate, showing that faithful tool use is not saturated. More importantly, models with similar aggregate scores fail in different ways: most stay disciplined on no-tool controls, while Llama-3.1 models show an Always-Call pattern, and at the same parameter scale Llama-3.1-70B and Qwen2.5-72B differ by 89 percentage points on control-task accuracy. Tool-use evaluation should measure not only whether
agents call tools, but whether they use tool outputs correctly and avoid tools when none is needed.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.RO
Do Vision-Language-Action Models Mean What They Say? On the Role of Faithfulness in Embodied Reasoning
Matthew Foutter, Matteo Cercola, Lena Wild et al. (+4 more)
Embodied Chain-of-Thought has emerged as a promising mechanism to enhance
robot decision-making and interpretability in black-box
Vision-
Language Action (VLA) models. However, whether this verbalized Chain-of-Thought truthfully reflects the policy's underlying decision process remains poorly understood. We distinguish between functional
reasoning, in which
reasoning improves task performance, and faithful
reasoning, in which
reasoning truly reflects the policy's internal decision process. We argue that SoTA
alignment strategies offer a necessary but insufficient notion of faithfulness, admitting
reasoning whose intermediate steps can mask the causal links in action prediction through confounding factors (e.g.,
reasoning that is ungrounded in the environment and internally disconnected or inconsistent), restricting policy generalization. We study this gap through a human evaluation of a SoTA
reasoning model for autonomous driving, revealing an inconsistent coupling between
reasoning quality and downstream trajectory improvement. We then operationalize a behavioral surrogate for
embodied faithfulness through a learned critic, Pinocchio, scoring observation grounding and stepwise coherence, and use this critic as a dense reward signal in post-training an
embodied policy with reinforcement learning. Across withheld driving
benchmarks, our post-trained planner improves faithfulness by 4% and 18% over SoTA
alignment and trajectory error post-training baselines, respectively, while maintaining competitive downstream task performance. Finally, on a synthetic out-of-distribution test set, post-training for faithfulness improves policy responsiveness to rare counterfactual scenarios by 1.6x that of a SoTA policy, suggesting that faithful
reasoning traces contribute to more robust, generalizable, and interpretable
embodied intelligence. Project page: https://mjf-su.github.io/pinocchio/
Read the full abstract ▾
All authors Matthew Foutter, Matteo Cercola, Lena Wild, Yunshan Wang, Michelle Li, Daniele Gammelli, Marco Pavone
Artificial Intelligence · Cross-listed from cs.OS
Elastic Gang: Per-Token Membership Change for a Hard-Barriered LLM Inference Gang Co-Scheduled with OS Processes
Daeyeon Son
On-device LLM decoding is a hard-barriered CPU-SIMD computation that wants every core for milliseconds per
token, while the rest of the OS wants those same cores continuously. A barriered gang cannot simply be dropped into a preemptive scheduler: an unannounced departure deadlocks a barrier, and an unannounced arrival silently corrupts logits. I present the elastic gang of Anima OS, a bare-metal x86-64 Rust kernel in which the inference gang is a first-class schedulable entity whose core membership may change between any two
tokens. The core mechanism is an ACK-latched epoch protocol that never waits on a named core: a seqlock-style generation-tagged latch composed with RCU/epoch-style membership consent, so each
token's participant set is the intersection of the cores the gang requested and the cores that acked the current epoch. An un-acked core is outside this
token and joins at most one
token later. Displaced general processes migrate and keep running; cores return to them the moment a generation ends. On a real AMD Zen 5 machine (8C/16T), inference output is bit-exact under verified per-
token membership change on both a 135M and a 7B model, the property that makes elasticity safe in a kernel whose safety gate reads logits. Against fair static core partitions, elastic membership Pareto-dominates: at intermediate inference duty cycles it delivers 1.75x (25%), 1.52x (50%), and 1.28x (75%) the general throughput of a static 8-core split at equal or better inference throughput, recovers all eight stranded cores when inference is idle, and converges to the split at saturation. Returning a lent core costs 0.22 us (p50); acquiring a busy, tenant-occupied core costs one scheduling
quantum (~16 ms): a running tenant is never preempted mid-slice. Decode throughput saturates at gang width 8, so ceding cores past the knee is nearly free: elasticity auto-sizes the gang online.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.CV
Targeted Structure Completion for Sparse-View 3D Reconstruction in Autonomous Driving
Guoqing Wang, Pin Tang, Xiangxuan Ren et al. (+2 more)
Reconstructing 3D scene structures from sparse, low-overlap observations remains a fundamental challenge in autonomous driving. Recent state-of-the-art frameworks achieve promising results by incorporating voxel-based Gaussians, but incur substantial computational redundancy due to a uniform volumetric processing strategy. To bridge the gap between the efficiency of pixel-based Gaussian methods and the structural completeness of voxel-based Gaussian approaches, we propose FocusGS, a simple yet effective framework that shifts the paradigm from global densification to targeted structural completion. Our central insight is that structural completion should be decoupled from deterministic regions, with computation concentrated exclusively on areas exhibiting
geometric ambiguity. Specifically, FocusGS addresses the localization challenge by deriving a 3D
Geometric Ambiguity
Manifold to accurately isolate localized areas prone to occlusion and high
geometric uncertainty. To overcome the subsequent
manifold completion challenge, we design a lightweight targeted structure completion module that selectively instantiates and optimizes continuous Gaussian queries strictly within this unstructured, sparse topological subspace. Extensive experiments demonstrate that FocusGS achieves a superior efficiency-quality trade-off, advancing state-of-the-art performance on driving-centric
benchmarks while naturally reducing the total number of Gaussians by ~74% and decreasing rendering time by ~34%.
Read the full abstract ▾
All authors Guoqing Wang, Pin Tang, Xiangxuan Ren, Liping Hou, Chao Ma
Artificial Intelligence · Cross-listed from cs.LG
Machine Learning for Depression Screening and Intervention: an Original Circadian Rhythm Score-based Methodology
Bin Wang, Shuo Lian, Yuanyuan Hou et al. (+5 more)
Depression screening from large-scale behavioral data is challenged by fragmented circadian indicators, limited interpretability, and the lack of intervention-oriented analysis. Existing approaches typically analyze sleep, activity, and social behaviors in isolation, failing to capture their joint circadian structure. To address this limitation, we first propose the Circadian Rhythm Score (CRS), a composite index that compresses multi-domain daily behaviors into a unified representation of circadian rhythm. CRS is constructed to maximize discriminative power for depression screening while preserving behavioral semantics through non-negativity constraints. Empirical results demonstrate near-lossless compression, where a single CRS retains almost the full predictive capability compared with multiple raw behavioral indicators. Building upon CRS, we develop an interpretable depression screening framework based on gradient-boosted trees and SHAP analysis, revealing nonlinear and saturation-like associations between circadian rhythm and depression risk. Beyond risk prediction, we further integrate interaction modeling and counterfactual regression to estimate heterogeneous and dose-dependent behavioral effects, enabling intervention-oriented
reasoning under different circadian contexts. Experiments on the China Health and Retirement Longitudinal Study (CHARLS, n=15,233), demonstrate robust screening performance (ROC-AUC=0.825) and identify actionable behavioral thresholds, including a minimum effective exercise dose of approximately 300 MET-min/week and an optimal restorative nap duration of approximately 65 minutes for sleep-deprived individuals. By bridging supervised representation learning and interpretable modeling, this work provides a scalable framework for depression screening and intervention-aware healthcare data mining.
Read the full abstract ▾
All authors Bin Wang, Shuo Lian, Yuanyuan Hou, Dexian Wang, Peilan He, Feng Hong, Yanwei Yu, Tianrui Li
Artificial Intelligence · Cross-listed from cs.CL
Retroactive Chain-of-Thought (RetroCoT): Forensic Reconstruction Prompts as a Safety Diagnostic Across Model Generations
Samira Hajizadeh
Safety
alignment in large
language models is typically evaluated against direct, imperative harmful requests. We show that this
alignment is highly conditioned on pragmatic register: models that refuse a direct request frequently comply when the same underlying objective is expressed through a different communicative stance. This suggests that current
alignment policies are not invariant to semantic equivalence, but remain sensitive to how a request is pragmatically framed. We introduce Retroactive Chain-of-Thought (RetroCoT), a single-turn attack that reframes harmful requests as forensic reconstruction tasks. Rather than requesting harmful instructions directly, RetroCoT presupposes that the harmful outcome has already occurred and asks the model, acting as a forensic analyst, to reconstruct in reverse the causal chain that produced it. On AdvBench (n=50), RetroCoT achieves attach success rate of 58% on gpt-4o and 52% on gpt-4o-mini, compared with direct-request baselines of 0% and 4%, respectively. We further identify a pronounced generation gap: GPT-5-family models refuse RetroCoT entirely, explicitly identifying the reconstruction premise in their refusal rationales, consistent with explicit coverage of this reconstruction register. However, this robustness does not generalize across pragmatic forms. A single
adversarial feedback turn presenting an existing forensic reconstruction response alongside evaluator critique raises ASR from 0% to 48% on GPT-5.4-mini and from 58% to 94% on GPT-4o; a control condition omitting the fabricated low score achieves 85% on GPT-5.4-mini, indicating that the operative element is pragmatic continuation within the established forensic frame rather than score manipulation. These results suggest that frontier-model
alignment remains conditioned on pragmatic framing rather than semantic intent, and that new pragmatic registers can continue to expose a...
Read the full abstract ▾
Artificial Intelligence
Formal Disco: Scalable Open-Ended Generation of Formally Verified Programs
Gabriel Poesia, Simon Henniger, Tzu-Han Hsu et al. (+2 more)
The cost of producing code is rapidly diminishing with increasingly capable AI
agents, while quality assurance of generated programs has not kept pace. Formal verification provides the strongest possible guarantees, but the ability of AI models to work with verification-aware
languages is hindered by the scarcity of human-written examples of programs in those
languages. To tackle this prevalent data scarcity issue, we propose Formal Disco: a distributed system for coordination of LLM-based workers that can be easily applied to open-ended synthetic data generation at scale. We use Formal Disco to share tasks and programs between three classes of workers: "initiators", which read random READMEs from open-source repositories and documentation snippets to sketch a related verified program, "fixers" which take compiler and verifier feedback and attempt to resolve issues, and "extenders" that take working programs and propose patches to expand them. Formal Disco records all
agent-generated traces and uses them both for initial
distillation from a stronger model as well as self-improvement. We also propose a principle of maximum
entropy for synthetic program generation, and use
entropy maximization via iterative supervised fine-tuning to learn to generate increasingly diverse programs over time. We release large
datasets of synthetic verified programs in three
languages - Dafny, Verus, and Frama-C -, and fine-tune open models for verification-relevant tasks, often matching or exceeding the performance of Claude Opus 4.5. Overall, our work offers a path to create synthetic data at scale for formal
reasoning domains and overcome the long-standing data barrier.
Read the full abstract ▾
All authors Gabriel Poesia, Simon Henniger, Tzu-Han Hsu, Yilun Du, Nada Amin
Artificial Intelligence · Cross-listed from cs.CV
Hierarchical Evidence-Driven Reasoning for Long Document Understanding
Junyu Xiong, Yonghui Wang, Rongjian Gu et al. (+4 more)
Retrieval-Augmented Generation (RAG) streamlines long-document understanding by leveraging
retrieval mechanisms to restrict input images to a highly curated subset. However, existing
multimodal RAG pipelines primarily face two critical challenges: first, standard semantic similarity retrievers frequently fetch topically overlapping yet answer-void distractor pages that mislead downstream generation; second, rigid single-pass pipelines heavily depend on initial
retrieval success, where any omission of core evidence inevitably causes cascading errors. To address these challenges, we introduce HIEVI-RAG, a hierarchical, evidence-driven
multimodal RAG framework for closed-domain document understanding. HIEVI-RAG systematically factorizes complex queries into a cooperative four-stage pipeline: (1) hierarchical question decomposition to break multi-hop root queries into atomic child questions; (2) coarse visual page
retrieval leveraging a
multimodal retriever to fetch candidate pages based on semantic similarity; (3) fine-grained page verification via EVIAGENT, a specialized multi-page verifier trained with GRPO to execute cross-page
reasoning over multi-image blocks; and (4)
memory-guided iterative generation that leverages accumulated sub-question context to execute multi-round, dynamic
reasoning over the prioritized sequence. Extensive evaluations across four
benchmarks demonstrate the robust efficacy and synergy of our framework, which significantly outperforms existing open-source baselines and exceeds the strongest reported baseline by an average of 8.05% in accuracy.
Read the full abstract ▾
All authors Junyu Xiong, Yonghui Wang, Rongjian Gu, Chenyu Liu, Bing Yin, Wengang Zhou, Houqiang Li
Artificial Intelligence
MRMS: A Multi-Resolution Memory Substrate for Long-Lived AI Agents
Jizhizi Li and Amy Shi-Nash
Long-lived AI
agents require continuity across interactions, but continuity cannot be obtained by simply extending the
prompt window. An
agent must preserve useful prior experience, retrieve it selectively, distinguish personal context from external evidence, and revise
memory when the underlying situation changes. We propose an architectural
memory substrate organized along two orthogonal axes: a representational axis spanning structured records, vector representations, and graph relations; and a temporal axis spanning short-term traces, medium-term abstractions, and long-term semantic commitments. Its key design constraint is synchronized structured-vector-graph
memory: structured records govern eligibility, vector representations support recall, and graph relations adjudicate support, contradiction, and supersession before gated context projection. Its central claim is that reliable personalization is a
memory design problem: useful
memory is structured, selectively exposed, continuously consolidated, and epistemically labeled rather than stored as undifferentiated conversation history. Beyond the framework, we instantiate MRMS as a lightweight prototype implementing structured records, vector
retrieval, temporal policies, and graph-based revision. The prototype exercises the core substrate mechanisms through pre-generation
memory selection, revision, boundary enforcement, and evidence attribution under controlled long-lived interaction scenarios with explicit evidence requirements.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.RO
SILO: Simulation-in-the-Loop Sim-to-Real Transfer for Multi-Stage Cable Routing
Stone Tao, Jie Xu, Hesam Rabeti et al. (+3 more)
Linear-deformable manipulation remains challenging due to the complex deformations of objects such as cables and ropes. Prior data-driven approaches, particularly imitation learning, have shown some promise in narrowly defined settings but typically require thousands of demonstrations for specific tasks and cable types, limiting scalability and generalization. We introduce a sim-to-real reinforcement learning (RL) framework for multi-stage cable routing that leverages GPU-parallelized simulation to approximate linear deformable behaviors. Training across thousands of parallel simulations enables the learned policies to generalize across diverse cable geometries and deformation patterns. To bridge the sim-to-real gap, we propose a novel deployment strategy that combines a Simulation In the LOop (SILO) execution framework, localized RL policies, and robust cable state estimation. On real-world cable routing tasks, our approach achieves higher success rates and 2x reduction in cycle times compared to prior state-of-the-art learning methods. To our knowledge, this is the first successful sim-to-real transfer of RL policies for multi-stage cable routing.
Videos and additional visualizations are available at https://silo-cable-routing.github.io/
Read the full abstract ▾
All authors Stone Tao, Jie Xu, Hesam Rabeti, Yashraj Narang, Yijie Guo, Iretiayo Akinola
Artificial Intelligence
Governed Individuation: Cryptographically Decoupling an Agent's Learning from Its Authority
Xue Qin, Simin Luan, Cong Yang et al. (+1 more)
Autonomous
agents are moving from sandboxed text generators to operators of code, data, and physical infrastructure, and they increasingly learn while deployed. This reopens a question that
alignment techniques answer only probabilistically: after an
agent has adapted in the field, is the running system still confined to what its operator authorised? Here we show that confinement can be guaranteed as an invariant of the
agent's execution architecture rather than a probabilistic outcome of its training. Governed individuation binds an
agent at boot to a cryptographically frozen identity digest, and routes every action through a gate defined over the semantic effect of the action rather than its name. We prove that no amount of learning, skill acquisition, or self-induced governance abstraction can widen the
agent's permitted authority without an operator-signed change to its identity; the guarantee holds even when the
agent induces its own safety principle and that principle is wrong. Empirically, in an open-ended tool-use
benchmark where a large action space rules out name-based blocking, ungoverned software
agents under reward pressure attempt to tamper with their own evaluation at a task-dependent rate that reaches every run on the hardest task, whereas the gate reduces executed forbidden effects to zero as a verified property of the construction while preserving task success. An
adversarial evaluation of monitors of increasing semantic depth shows false-allows falling from 75% (name-based gating) to zero (dynamic effect tracing), and refusal history transfers compliance to held-out red-line families. Trust in a deployed learning
agent shifts from a wager on its continued
alignment to a check anyone can run at boot.
Read the full abstract ▾
All authors Xue Qin, Simin Luan, Cong Yang, Zhijun Li
Artificial Intelligence · Cross-listed from cs.CV
G2VD: Generalizable AI-Generated Video Detection via Counterfactual Intervention and Causal Disentanglement
Meng Du, Hongchang Chen, Ran Li et al. (+4 more)
Rapid advances in AI
video generation pose increasing security risks and call for reliable detectors with strong cross-domain generalization. Although existing methods perform well under in-domain evaluation, their performance degrades substantially on unseen generators. A key reason is shortcut learning, where detectors rely on domain-specific bias rather than intrinsic forensic cues. To address this issue, we propose G2VD, a generalizable AI-generated
video detection framework based on counterfactual intervention and causal disentanglement. First, G2VD introduces a counterfactual intervention pipeline (CFIPipeline) that constructs counterfactual samples through VAE-based reconstruction and subsequent frequency-domain and pixel-domain
alignment, thereby weakening spurious correlations between domain-specific bias and authenticity labels. Building on this intervention, we further design a causal disentanglement classifier that combines two domain-anchored branches with complementary objectives and a constraint based on the
Hilbert-Schmidt Independence Criterion (HSIC), encouraging the causal and non-causal representations to capture intrinsic forensic cues and domain-specific bias, respectively. Experiments across four public
datasets demonstrate strong cross-domain performance and consistent gains over baseline methods. In the challenging GenVidBench setting, G2VD achieves over 90% overall ACC, with improvements of 0.194 in F1 and 0.104 in AUC over comparable state-of-the-art methods, while using only 10% of the available training data. Code is available at https://github.com/DMOSCAR-98/G2VD.
Read the full abstract ▾
All authors Meng Du, Hongchang Chen, Ran Li, Junjie Zhang, Qi Ouyang, Shibo Zhang, Shuxin Liu
Artificial Intelligence · Cross-listed from cs.IR
Do All Visual Tokens Matter Equally? Object-Evidence Preserving Token Merging for Vision-Language Retrieval
Suhyeong Park, Junha Jung, Jungwoo Park et al. (+1 more)
Multi-vector
vision-
language retrieval preserves fine-grained visual evidence through maximum-similarity late interaction, but dense image-side
tokens make storage and scoring expensive. Existing
token compression methods reduce this cost, yet they can remove or collapse object- and region-level evidence that future query
tokens may need to select. We propose SaMer, an object-aware
token merging framework that compresses image-side post-projector
tokens into $K$ representative centroids while preserving the original late-interaction interface. SaMer uses object annotations only during training as a merge prior to discourage cross-instance mixing, requires no ground-truth bounding boxes or detectors at inference time, and adapts only the shared projection layer with frozen
vision and
language backbones. With $K=64$, SaMer removes more than 93% of image-side
tokens and reduces ColPali storage by $16.09\times$, while improving R@1 on Flickr30K and MSCOCO. These gains arise because object-aware merging preserves query-selectable object evidence that
pruning or feature-only pooling can remove or collapse. SaMer also outperforms compression baselines and shows stronger phrase-level grounding, suggesting that efficient multi-vector
retrieval depends not only on reducing
token count, but on preserving the evidence future query
tokens need to select.
Read the full abstract ▾
All authors Suhyeong Park, Junha Jung, Jungwoo Park, Jaewoo Kang
Artificial Intelligence · Cross-listed from cs.CV
LCPNet: Latent Consistent Proximal Unfolding Network for Infrared Small Target Detection
Tianfang Zhang, Lei Li, Chang Liu et al. (+3 more)
Infrared small target detection (IRSTD) aims to identify long distance small targets from complex infrared backgrounds, and is a fundamental task in remote sensing. Deep learning methods have improved IRSTD by learning discriminative image-to-mask mappings, but such feed-forward designs often underuse physical decomposition structure between targets and backgrounds. Deep unfolding methods partially address this issue by
embedding model-driven iterations into neural networks, yet existing designs still operate mainly in image domain and use updates and
memory mechanisms that are not fully coupled with underlying optimization process. To address these limitations, we propose Latent Consistent Proximal unfolding network (LCPNet). First, we verify that low-rank prior remains valid in latent representations and perform unfolding in this space, preserving physical constraint while avoiding repeated compression of intermediate states. Second, we derive a Latent Consistent Proximal (LCP) solver that evolves each latent variable from its previous state rather than reconstructing through an indirect residual, and stabilizes small target updates through task-adaptive normalization and gain control. Third, we introduce Shared Optimization
Memory (SOM), a common historical state shared by all decomposition variables to provide coordinated guidance across unfolding stages. Extensive experiments on four public
benchmarks demonstrate that LCPNet achieves accurate and robust detection with low false-alarm rates, together with competitive efficiency among high-accuracy deep unfolding methods. Model and code are available at Model and code are available at https://github.com/Tianfang-Zhang/LCPNet.
Read the full abstract ▾
All authors Tianfang Zhang, Lei Li, Chang Liu, Zhenming Peng, Huaping Zhang, Xiangyang Ji
Artificial Intelligence · Cross-listed from cs.CV
TORINO: Token Reduction via Interpretable Concept Overlap in Vision-Language Models
Riccardo Renzulli, Gabriele Spadaro, Shruthi Gowda et al. (+2 more)
Vision-
Language Models (VLMs) have demonstrated impressive capabilities across different tasks, but their computational cost is dominated by the large number of visual
tokens fed to the
language model. Existing
token reduction methods rely on
attention-based scores or pairwise similarity, without an explicit semantic representation of each
token. We introduce TORINO (
TOken Reduction via Interpretable coNcept Overlap), a plug-and-play framework for adaptive visual
token reduction in VLMs that requires no fine-tuning of the underlying model. TORINO leverages Sparse Autoencoders (SAEs) to project visual
tokens into an interpretable latent space where
token relationships can be analyzed through shared concept activations. Specifically, we define concept overlap as the degree of agreement between active SAE latents and use it to group
tokens that share semantic content. Reduction within each group is then performed by either
pruning or merging, providing a unified framework that preserves semantically important visual information while removing redundancy. Unlike fixed-budget approaches, TORINO dynamically adapts the reduction rate to input complexity, allowing different images to retain different numbers of
tokens. Experiments across multiple
vision-
language benchmarks show that TORINO achieves favorable efficiency-accuracy trade-offs, reducing the number of visual
tokens with minimal performance loss. Source code released at https://eidoslab.github.io/TORINO.
Read the full abstract ▾
All authors Riccardo Renzulli, Gabriele Spadaro, Shruthi Gowda, Alaa Eddine Mazouz, Van-Tam Nguyen
Artificial Intelligence · Cross-listed from cs.RO
Simple-to-Complex Structured Demonstrations for Vision-Language-Action Learning
Xinchuan Qiu and Yi Yu
Vision-
Language-Action (VLA) models have demonstrated strong capabilities in robotic manipulation by integrating visual perception,
language understanding, and
robot action generation. Existing research has primarily focused on improving model architectures, training strategies, and
dataset scale, while little
attention has been paid to how demonstrations are collected and organized. We identify demonstration organization as a fundamental yet overlooked aspect of imitation learning, as it directly affects policy learning efficiency, training stability, and policy generalization. To address this gap, we propose a simple-to-complex structured demonstration collection strategy for VLA learning using a dual-arm robotic platform. Our approach systematically organizes data through three general principles: (i) decomposing complex manipulation tasks into progressively learnable sub-skills, (ii) standardizing the interaction environment to reduce unnecessary variability, and (iii) organizing demonstrations according to progressively increasing task complexity. This structured design enables VLA models to first acquire fundamental manipulation skills before learning increasingly complex task compositions, facilitating more effective learning of long-horizon manipulation tasks. We evaluate the proposed strategy on two representative robotic manipulation tasks: block grasping and sorting, and towel folding. Experimental results show consistent improvements in task success rate and training stability compared with the baseline method of directly collecting end-to-end complete task trajectories. These findings highlight demonstration organization as a previously underexplored but important factor in VLA learning and provide practical insights into efficient skill acquisition, scalable
dataset construction, and long-horizon robotic manipulation.
Read the full abstract ▾
Artificial Intelligence
Attention Limited Reward Learning
Wenqian Xing
Pairwise human comparisons are a primary interface through which modern AI systems learn human preferences. RLHF and related
alignment pipelines typically model such comparisons with Bradley–Terry log-odds, where choice probabilities are governed by latent reward differences. This paper examines what this assumption misses through a reduced-form model motivated by rational inattention, in which each label is generated by a low-capacity evaluation channel. The model separates two forms of ambiguity that standard reward modeling tends to conflate: a comparison may be difficult because the two candidates are genuinely close in value, or because the relevant distinction is hard to detect under limited
attention. We show that limited
attention can fundamentally distort what pairwise comparisons reveal. In particular, passive comparison data cannot generally distinguish reward,
attention, and default tendencies, and heterogeneous
attention can make standard Bradley–Terry reward modeling recover misleading rankings. Our analysis shows that learning is governed not by the raw number of labels, but by the amount of attended information each label carries. A case study on human votes over
language-model pairs from Chatbot Arena exhibits the predicted signature, a cyclic component of the comparison data that exceeds sampling noise and that no scalar reward can represent; a second case study on perceptual comparisons shows that response times and gaze carry gap information that the labels do not. This perspective suggests that human feedback should be treated not as direct revealed preference, but as an
attention-limited measurement process: a weak preference signal may reflect hidden evaluation difficulty rather than genuine indifference.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.SE
LLM-Driven CI-CD Workflow Intelligence for Cyber Systems Engineering
Bonan Shen, Jiazhou Gao, Tao Ning et al. (+2 more)
CI/CD workflows have become executable operational policy: they decide what gets built, tested, released, and deployed, and they mediate how maintainers interact with delivery infrastructure. That makes them an important measurement point for cyber-systems engineering. Recent large
language model (LLM) work shows that workflow stages can be recognized directly from configuration files, but stage labels alone do not tell us whether a workflow is brittle, unusual for its ecosystem, or worth revising first. We present an LLM-based CI/CD analysis pipeline that combines repository enrichment, anti-pattern detection, stage mining, and recommendation generation over a large GitHub corpus. Starting from 59,550 repositories with at least 1,000
stars, we identify 34,225 projects with CI/CD and collect 127,559 configuration files. Across 75,201 analyzed workflows, the anti-pattern detector reports 434,769 findings, dominated by reliability and maintainability issues. Across 59,906 configurations, stage usage differs significantly by
language ($\chi^2 = 4168.88$, $p < 0.001$, Cramer's $V = 0.063$), and domain analysis shows distinct operational profiles, including higher release and cache usage in mobile projects. For repository-level recommendation generation, few-shot prompting performs best overall, averaging 8.25 recommendations per repository with 96.1% YAML-valid snippets. Taken together, the results argue for CI/CD observability that combines diagnosis, context, and human review rather than treating workflow mining as a stage-classification problem alone.
Read the full abstract ▾
All authors Bonan Shen, Jiazhou Gao, Tao Ning, Wei-Jung Huang, Xin Liu
Artificial Intelligence · Cross-listed from cs.LG
A Few Teacher Steps Go a Long Way: Cost-Efficient On-Policy Data Augmentation for Agent Post-Training
Junze Ye, Jiayi Cheng, Miao Lu et al. (+3 more)
For LLM
agents, supervised fine-tuning is not only about teacher labels' quality, but also about which interaction contexts those labels condition on. Pure behavioral cloning uses full teacher demonstrations, creating a mismatch between teacher-induced contexts seen in training and student-induced contexts encountered at test time. Recent work addresses this mismatch by querying a teacher at contexts reached by the student, often with increasingly elaborate filtering of the teacher's continuations. We instead frame on-policy data construction as a budget-allocation problem: under matched supervision resources, should teacher output be spent on more start-to-finish demos, longer continuations, outcome filtering, or broader coverage of learner-induced contexts? We formalize this design space through the rollout policy, switch-time distribution, continuation horizon, filtering rules, and two complementary costs: teacher inference generated before filtering and teacher supervision retained for SFT. Across HotpotQA, ALFWorld, and Terminal-Bench-Dev, bounded unfiltered teacher continuations at learner-induced contexts improve over pure behavioral cloning at matched budgets. On HotpotQA and ALFWorld, where we run the full comparison, few-step continuations match or exceed success-filtered and critical-context-filtered alternatives. Our findings suggest that a few teacher steps, placed at learner-induced contexts, can be a more cost-efficient supervision allocation than longer or more heavily curated teacher completions.
Read the full abstract ▾
All authors Junze Ye, Jiayi Cheng, Miao Lu, Michal Mankowski, Jose Blanchet, Mohsen Bayati
Artificial Intelligence
Context-Masked Truncated Reasoning Audits for Answer-Key Dependence in LLM Tutors
Bonan Shen, Dingyan Shang, Youting Wang et al. (+2 more)
Large
language model (LLM) tutors may have access to teacher notes, answer keys, rubrics, or retrieved solutions while producing student-facing explanations. We study whether truncated
reasoning probes can distinguish direct access to such private context from answer information carried by the written explanation. Using Truncated
Reasoning AUC Evaluation (TRACE), we evaluate 1000 GSM8K problems under question-only, correct answer-key, and wrong answer-key contexts. When forced-answer probes retain the private key, answer-key TRACE AUC rises from 0.375 to 0.900, and the gold answer is recoverable with no explanation at all in 998 of 1000 cases. We then introduce a context-masked replay: answer-key-generated prefixes are probed under the corresponding question-only
prompt. Masking reduces 10% prefix accuracy from 0.997 to 0.126 and median AUC from 0.900 to 0.375, nearly matching question-only values of 0.113 and 0.375. On 746 pairs where both explanations end correctly, the masked mean AUC difference is $-0.0086$ with a 95% bootstrap interval spanning zero. Wrong keys still account for 272 of 387 incorrect final responses, showing that private artifacts can influence outputs even when early-prefix evidence disappears after masking. These results establish context masking as necessary for attributing early answer availability to an explanation rather than its hidden input.
Read the full abstract ▾
All authors Bonan Shen, Dingyan Shang, Youting Wang, Tao Ning, Bowen Liu