4 papers
ASCT: Attentive Search over Counterfactual Trees for Credit Assignment in Agentic Reinforcement Learning
Yang Li, Jinhan Yang, hai liu +8
Terminal utility evaluates a complete agentic workflow, but learning requires credit for the decisions within it. We introduce Attentive Search over Counterfactual Trees (ASCT), a…
From Evidence to Effect: Authority Semantics and Runtime Infrastructure for Stateful Agents
Yang Li, Sergey Volkov, Hai Liu +9
Stateful agents reuse artifacts after producing executions and permissions change. We formalize authority-sufficient observations and durable effects bound to execution and materia…
Agent-UCT: Upper Confidence Bounds Applied to Trees for Cost-Aware Agentic Workflow Optimization
Yang Li, Hai Liu, Dian Shao +9
Workflow search pays for complete quality observations, whose execution cost changes as reusable prefixes accumulate. AgentUCT uses an economic model of the remaining evaluation co…
MAPL: Multi-Objective Preference Learning for Robot Locomotion
Xiyue Chen, Muhan Lin, Shuyang Shi +1
Reward design remains a major bottleneck in reinforcement learning for robot locomotion, where successful policies often depend on carefully tuned, task-specific reward functions.…