Recursive self-improvement (RSI) envisions agent systems that autonomously improve their own models and harnesses, with each updated system continuing to improve itself. Existing approaches typically restrict optimization objectives, editable components, or evaluation procedures, limiting open-ended exploration and con...
Wan-Ying Qu, Zhi-Yuan Fan, Ji-Yao Liu et al.· 0 citations
It is demonstrated that MCP tool-call traces can act as effective behavioral side channels for user profiling, motivating the development of privacy-aware mechanisms for agent-tool interactions.
Dong-Hang Duan, Qing-Yu Liu, Yi-Fei Liu et al.· 0 citations
AI agents can inspect code and run tests to support smart contract security. SlowMist reports about $38.6 billion in losses across the broader blockchain ecosystem. Existing evaluations can reward memorized or copied answers, including attack code recovered from public blockchain transactions. Reconstructing contracts...
Zi-Fan Peng, Peng Huang, Jing-Yi Zheng et al.· 0 citations
It is observed that VLMs can clearly distinguish benign inputs from harmful ones in their representation space, and even among harmful inputs, jailbreak samples form a distinct internal state that is separable from refusal samples.
Zhi-Hua Wei, Qiang Li, J. Ruan et al.· arXiv.org· 6 citations
Large language model (LLM)-based agents can exhibit deceptive behavior during task execution, including hiding failures, fabricating results, or falsely signaling task completion. Existing monitoring approaches mainly detect deception after it appears in observable actions or outputs. In this paper, we investigate whet...
Xin-Ling Li, Da-Di Guo, Qing-Yu Liu et al.· 0 citations
This work introduces a multi-agent framework designed to perform problem evolution while validating the solvability and increased difficulty of the generated problems, and provides empirical evidence that code-driven agents can serve as a viable mechanism for synthesizing high-difficulty mathematical reasoning problems...
Dadi Guo, Yue-Jin Xie, Qing-Yu Liu et al.· 0 citations
This work identifies and formalizes this failure mode, which is term destructive resource preemption: obtaining the resources required for a requested task by terminating, overwriting, evicting, or degrading an incumbent task.
Through harness-policy co-evolution, SafeEvolve converts safety experience into an evolved runtime harness and improved policy behavior, and experiments show that SafeEvolve achieves a stronger safety-utility tradeoff than existing baselines.
Safety Harness Evolution (SHE) is proposed, a framework that learns evolving safe boundaries from rollout trajectories and introduces an attribution-guided evolution loop that converts trajectory failures into structured diagnoses, learns artifact-specific boundary refinements, and selects evolved harnesses through saf...
Wanying Qu, Qing-Hua Mao, Yu Li et al.· 5 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.