AlphaZero Auxiliary Loss improves play in sparsely rewarded games
A new paper studies the gap between strong and perfect play in AlphaZero for Connect Four and Chomp, introducing AlphaZero Auxiliary Loss (AZAL) that adds oracle-derived policy supervision. The work highlights limits of vanilla AlphaZero in sparsely rewarded games and shows auxiliary supervision can narrow the gap.
Entities
Related
Paper proposes correlation-aware bandits with surrogate rewards for LLM routing
A new arXiv paper introduces contextual bandit algorithms that leverage surrogate reward signals from machine learning models to improve LLM routing decisions. The approach accounts for inter-arm correlations and noisy auxiliary rewards, addressing limitations of classical bandits that assume conditional independence.
Agora paper proposes auction-based task allocation for LLM agents
A new arXiv paper introduces Agora, a framework that uses an incentive-compatible auction mechanism to dynamically allocate tasks to expert LLMs and tools, aiming to improve reasoning performance while accounting for cost and performance variability. The approach addresses limitations in current orchestration methods that rely on coarse-grained function matching.
LLM/VLA models enable prompt-driven exploration in RL
A new research paper proposes using large language models (LLMs) and vision-language-action (VLA) models to drive exploration in reinforcement learning by modifying natural language prompts, which induce global behavioral changes beyond standard action noise. This approach could help policies escape weak local optima more effectively.