Implicit Compression Regularization for Efficient RL Reasoning
Discover how Implicit Compression Regularization improves reinforcement learning by shortening reasoning length without losing accuracy.
Scale-Conditioned Evaluation of AI Agent Memory Usability
Discover how scale-conditioned evaluation improves AI agent memory by measuring reliability amid irrelevant data growth and optimizing retrieval performanc...
SOM: Enhanced Opponent Modeling for LLM Agents Using SCM
Discover SOM, a novel framework using Structural Causal Models to improve opponent prediction accuracy and stability in LLM-based agents.
Signal Reshaping for GRPO to Boost Weak-Feedback Code Repair
Enhance agentic code repair with signal reshaping in GRPO, improving accuracy and performance under weak feedback conditions in reinforcement learning.
Role-Aware Policy Optimization Boosts Multimodal Reasoning
Discover how Structured Role-Aware Policy Optimization improves multimodal reasoning in vision-language models by refining token-level credit assignments.