Xunliang Cai

Recursive belief updates in AgentOPSD reveal pivotal decision points, leading to a 89.1% success rate on complex RL tasks.

AFD-Ledger reveals that optimizing deployment for AFD can drastically cut evaluation costs while exposing the nuanced performance dynamics between homogeneous and heterogeneous setups.

Selective distillation can unlock critical learning signals in reinforcement learning, leading to significant performance gains in complex tasks.

SkillRise achieves up to 8.5 percentage points better performance than leading methods by effectively reusing transferable skills across related tasks.

Game solvers can teach LLMs how to make better decisions in long-horizon tasks by providing actionable turn-level feedback, leading to superior performance in complex environments.

ACE revolutionizes context management for LLM agents, enabling them to adaptively retain critical information without loss, leading to superior decision-making performance.

Current search agents fall short of user expectations, with a new benchmark revealing critical gaps in their performance on everyday tasks.

SIRI allows LLM agents to autonomously develop and internalize skills, achieving up to a 2.2% performance boost without external dependencies.

MLLMs excel at single-hop tasks but falter dramatically in open-world scenarios, revealing critical gaps in their reasoning capabilities.

LLM agents can internalize skills via in-context RL, achieving zero-shot autonomous behavior without the token overhead and retrieval noise of traditional methods.

Forget hand-tuning rollout budgets: $V_{0.5}$ dynamically allocates compute to sparse RL rollouts based on a real-time statistical test of a generalist value model's prior, slashing variance and boosting performance.