paper-with-me

홈 › Papers

Verifiability-First Agents: Provable Observability and Lightweight Audit Agents for Controlling Autonomous LLM Systems

2025-12-19 · Abhivansh Gupta arxiv

As LLM-based agents grow more autonomous and multi-modal, ensuring they remain controllable, auditable, and faithful to deployer intent becomes critical. Prior benchmarks measured the propensity for misaligned behavior and showed that agent personalities and tool access significantly influence misalignment. Building on these insights, we propose a Verifiability-First architecture that (1) integrates run-time attestations of agent actions using cryptographic and symbolic methods, (2) embeds lightweight Audit Agents that continuously verify intent versus behavior using constrained reasoning, and (3) enforces challenge-response attestation protocols for high-risk operations. We introduce OPERA (Observability, Provable Execution, Red-team, Attestation), a benchmark suite and evaluation protocol designed to measure (i) detectability of misalignment, (ii) time to detection under stealthy strategies, and (iii) resilience of verifiability mechanisms to adversarial prompt and persona injection. Our approach shifts the evaluation focus from how likely misalignment is to how quickly and reliably misalignment can be detected and remediated.

📄 PDF Abstract BibTeX arXiv:2512.17259

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Benchmarking Partial Observability in Reinforcement Learning with a Suite of Memory-Improvable Domains

2025-07-31 · Ruo Yu Tao, Kaicheng Guo, Cameron Allen, George Konidaris arxiv

Mitigating partial observability is a necessary but challenging task for general reinforcement learning algorithms. To improve an algorithm's ability to mitigate partial observability, researchers need comprehensive benc…

General Reinforcement Learning

Entropy-Based Observability for AI Agent Behavior

2026-06-04 · Olasimbo Ayodeji Arigbabu arxiv

AI agents are typically instrumented through outcome-oriented indicators such as task success, reward, latency, and cost.Although these indicators are operationally important, they provide limited visibility into the int…

Privacy-Preserving Verifiable Neural Network Inference Service

2024-11-12 · Arman Riasi, Jorge Guajardo, Thang Hoang

Machine learning has revolutionized data analysis and pattern recognition, but its resource-intensive training has limited accessibility. Machine Learning as a Service (MLaaS) simplifies this by enabling users to delegat…

Privacy Preserving

Forager: a lightweight testbed for continual learning with partial observability in RL

2026-05-01 · Steven Tang, Xinze Xiong, Anna Hakhverdyan, Andrew Patterson 외 arxiv

In continual reinforcement learning (CRL), good performance requires never-ending learning, acting, and exploration in a big, partially observable world. Most CRL experiments have focused on loss of plasticity -- the ina…

Reinforcement LearningContinual Learning

Towards Trustworthy Multi-Turn LLM Agents via Behavioral Guidance

2025-12-12 · Gonca Gürsun arxiv

Large Language Models demonstrate strong reasoning and generation abilities, yet their behavior in multi-turn tasks often lacks reliability and verifiability. We present a task completion framework that enables LLM-based…

Reinforcement Learning