Large language model agents are increasingly extended with reusable skills, but untrusted skill content can induce operations outside the user's task. Detecting this contextual risk requires more than suspicious-pattern matching, while dynamic analysis incurs the cost of skill execution. We find that hidden states formed during prefill contain readable information about task-operation consistency, including across held-out expression forms in controlled evaluations. Matched Operation-state interventions also affect later Boundary states and shift fixed risk readouts toward the donor label. Based on these findings, we introduce LatentTriad, which combines contextual Boundary features, group-adjusted Operation directions, and departures from benign state relationships. All branches share one prefill: detection generates no output tokens, invokes no tools, and executes no skill code. Across 5,830 skill packages from five sources and seven backbones, LatentTriad achieves 99.0-99.2% pooled F1 and 93.0-94.6% source-macro F1. A separately measured compact implementation completes prefill and detection in a median of 259-284 ms. These results connect pre-action risk readability and functional influence to execution-free skill screening.
Library fuzzing synthesizes harnesses that expose new API interactions, but deploying it end to end still requires extensive project-specific integration. Existing LLM-based systems automate individual steps, yet they treat execution outcomes as coarse signals: a stalled campaign does not reveal which state blocks progress, and a crash trace does not reveal whether the library or the harness is at fault. We present FuzzAgent, a multi-agent system that builds and evolves a complete fuzzing campaign from a source tree alone. Specialized agents acquire project knowledge and repair campaign artifacts automatically. Hierarchical coverage diagnosis localizes blockers from modules down to branches, and dependency-aware generation converts them into harnesses with the required API lifecycle. Crash validation recovers runtime object states to distinguish harness misuse from library bugs. We evaluate FuzzAgent on 20 real-world C/C++ libraries against five state-of-the-art baselines (OSS-Fuzz, OSS-Fuzz-Gen, PromptFuzz, PromeFuzz, and ProteusFuzz). FuzzAgent completes the full fuzzing lifecycle for all 20 libraries without human intervention at $3.17 per library. On the 20 libraries, it reaches 179619 branches, exceeding OSS-Fuzz, PromptFuzz, ProteusFuzz, PromeFuzz, and OSS-Fuzz-Gen by 45.1%, 73.2%, 87.7%, 92.1%, and 191.2%, respectively. FuzzAgent also identifies 102 new library bugs, 90 of which have already been acknowledged or fixed by upstream maintainers.
Tianjian Li, Chuanyang Jin, Seungone Kim, Tong Chen, John Hoffman, Yoram Bachrach amd Pedro Rodriguez, Ping Yu, Eryk Helenowski 等 13 位作者
Autoresearch agents can now attempt long-horizon AI research end-to-end, i.e., self-improvement. However, best results are still obtained by research agents interacting with human researchers, and improving AI's ability to perform this co-research will only reinforce this effect---a process termed co-improvement (Weston and Foerster, 2025). Further, aligning with human expectations is likely a safer path for research agents, especially given recent agent reward hacking phenomena (OpenAI, 2026b). In this work we address these issues in four ways: (i) we empirically study the role of humans in directing autoresearch agents, focusing on the domain of benchmark creation, and find humans still play a pivotal role particularly in the ideation phase; (ii) we also develop a benchmark, HackBench, that shows current autoresearch agents are not aligned and are unable to achieve hack localization sufficiently; (iii) we present a method to train agents for improved human interaction by using human simulators as a proxy during reinforcement learning (RL); and finally (iv) we develop a method to improve model text generation, and hence agent communication, by RL with rubrics aligned with human experts. Together, our training and evaluation recipes provide a path toward more capable and aligned (co-)research agents.
Shunwen Bai Zhejiang University, Jiahuan Zhang, Haoran Huang, Yurun Wang, Jiale Liu, Yanxi Wu Zhejiang University, Ningzhe Yu, Yudong Gao Hong Kong University of Science and Technology 等 9 位作者
Currently, large language models (LLMs) have significant limitations in spatial reasoning, particularly in the absence of visual input. To address this issue, we introduce SODA (Spatial OODA), which draws inspiration from the OODA cognitive loop (Observe, Orient, Decide, Act), originally designed to enhance human decision-making in dynamic environments. Specifically, we embed the OODA loop into multiple control tasks, generating the SPOD-143k dataset, and successfully integrate it into LLMs through a two-phase and spatia-aware training strategy (SFT and GRPO). Furthermore, to fill the gap in evaluating spatial reasoning in purely text-based LLMs, we introduce the SPOD-Bench benchmark, including multiple tasks divided into three levels of difficulty. Experimental results show that SODA significantly enhances the spatial reasoning capabilities of LLMs across testing scenarios including SPOD-Bench, SPACE and applications, providing a replicable and effective paradigm for improving the spatial cognition of LLMs.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表