相关论文: Asymmetric Goal Drift in Coding Agents Under Value…
Recent advances in learning techniques have garnered attention for their applicability to a diverse range of real-world sequential decision-making problems. Yet, many practical applications have critical constraints for operation in real…
Future 5G systems will need to support ultra-reliable low-latency communications scenarios. From a latency-reliability viewpoint, it is inefficient to rely on average utility-based system design. Therefore, we introduce the notion of…
Humans often become more self-aware under threat, yet can lose self-awareness when absorbed in a task; we hypothesize that language models exhibit environment-dependent \textit{evaluation awareness}. This raises concerns that models could…
Value alignment problems arise in scenarios where the specified objectives of an AI agent don't match the true underlying objective of its users. The problem has been widely argued to be one of the central safety problems in AI.…
Large language model (LLM) agents are fundamentally constrained by context length on long-horizon tasks. We introduce Context-Folding, a framework that empowers agents to actively manage their working context. An agent can procedurally…
Frontier AI regulations primarily focus on systems deployed to external users, where deployment is more visible and subject to outside scrutiny. However, high-stakes applications can occur internally when companies deploy highly capable…
Large language model (LLM)-based agents are increasingly used to perform complex, multi-step workflows in regulated settings such as compliance and due diligence. However, many agentic architectures rely primarily on prompt engineering of a…
Many modern software systems are built as a set of autonomous software components (also called agents) that collaborate with each other and are situated in an environment. To keep these multiagent systems operational under abnormal…
As autonomous agents become more capable of performing real-world tasks, distinguishing scheming behavior from benign task pursuit may become a central AI control problem. Existing monitors often rely on chain-of-thought access or internal…
The rise of large language models for code has reshaped software development. Autonomous coding agents, able to create branches, open pull requests, and perform code reviews, now actively contribute to real-world projects. Their growing…
Conventional distributed approaches to coverage control may suffer from lack of convergence and poor performance, due to the fact that agents have limited information, especially in non-convex discrete environments. To address this issue,…
We systematically evaluate the quality of widely used adversarial safety datasets from two perspectives: in isolation and in practice. In isolation, we examine how well these datasets reflect real-world adversarial attacks based on three…
Coding agents now run autonomously with shell, file, and network privileges. When a user issues a benign request, the agent sometimes does more than asked: it deletes unrelated files, wipes a stale credentials backup, or rewrites…
AI code generation tools are transforming software development, especially for novice and non-software developers, by enabling them to write code and build applications faster and with little to no human intervention. Vibe coding is the…
Coding agents powered by LLMs face critical sustainability and scalability challenges in industrial deployment, with single runs consuming over 100k tokens and incurring environmental costs that may exceed optimization benefits. This paper…
Autonomous exploration in mobile robotics often involves a trade-off between two objectives: maximizing environmental coverage and minimizing the total path length. In the widely used information gain paradigm, exploration is guided by the…
This work presents a decentralized time synchronization algorithm for multi-agent systems. Each agent possesses two clocks, a hardware clock that is perturbed by environmental phenomena (e.g., temperature, humidity, pressure, g forces,…
Multiagent AI systems require consistent communication, but we lack methods to verify that agents share the same understanding of the terms used. Natural language is interpretable but vulnerable to semantic drift, while learned protocols…
We identify and formalize a novel security risk: Context-Fragmented Violations (CFVs) - a class of policy breaches where individual agent actions appear locally safe and reasonable, yet collectively violate organizational policies because…
Reward hacking--where agents exploit flaws in imperfect reward functions rather than performing tasks as intended--poses risks for AI alignment. Reward hacking has been observed in real training runs, with coding agents learning to…