Asymmetric Goal Drift in Coding Agents Under Value Conflict
Explore how coding agents face asymmetric goal drift under value conflicts, impacting security and privacy in autonomous systems.
Context-Sensitive Abstractions in RL with Parameterized Actions
Discover how context-sensitive abstractions enhance reinforcement learning with parameterized actions for improved AI decision-making and sample efficiency...
Auction-Based Method Boosts Language Agent Communication
Discover DALA, an auction-based framework that optimizes language agent communication by reducing token costs and enhancing efficiency.
Mitigating Self-Jailbreak in Large Reasoning Models Safely
Discover how Chain-of-Guardrail (CoG) improves safety in large reasoning models by preventing self-jailbreak without sacrificing reasoning power.
Test-Time Matching Boosts Compositional Reasoning in AI
Discover how Test-Time Matching and group matching score improve compositional reasoning in multimodal AI models, surpassing previous benchmarks.