arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2610.14636cs.AI
收藏

PEARL:面向多轮智能体的过程经验感知强化学习

PEARL: Process-Experience Aware Reinforcement Learning for Multi-Turn Agents

Fudan University(复旦大学);Shanghai Innovation Institute(上海创新研究院)

Jie Yang, Mianqiu Huang, Yiwei Zhao, Bo Wang, Yang Liu, Yuxin Wang

AI总结:

该研究针对多轮智能体训练中局部信用分配不足的问题,提出PEARL算法,通过模式记忆选择性更新策略,在AppWorld和ALFWorld数据集上取得优于基线的性能。

摘要:

终端奖励使多轮智能体无需步骤级标注即可训练,但组相对优势几乎无法指导调整哪些中间决策。失败轨迹中也包含合理动作,且不利决策模式会随训练推进反复出现。我们提出PEARL(Process-Experience Aware Reinforcement Learning,过程经验感知强化学习),无需额外评论家即可改进局部信用分配。给定同一历史中具有不同结果的两个后续序列,PEARL定位它们的首次分歧点,并构建每个分支局部决策的特征表示。它在近期失败与成功记忆中对这些决策特征进行聚类,以识别反复出现的局部模式。利用该模式记忆,PEARL选择性地在低奖励决策处强化负向优势,同时保持正向优势和基础策略优化器不变。在AppWorld上,PEARL在4B和8B规模下的普通任务与挑战任务均优于ARPO;在ALFWorld上,其在4B规模下达到97.9%的平均成功率,比最强基线高2.7个百分点,在8B规模下仍具竞争力。对塑形位置、成功库过滤及记忆窗口的消融实验,验证了基于反复模式的选择性更新的价值。

英文摘要:

Terminal rewards make multi-turn agents trainable without step-level annotations, but group-relative advantages provide little guidance about which intermediate decisions to adjust. Failed trajectories also contain reasonable actions, while unfavorable decision patterns can recur as training proceeds. We introduce PEARL (Process-Experience Aware Reinforcement Learning), which improves local credit assignment without an additional critic. Given two continuations from the same history with different outcomes, PEARL locates their first divergence and builds a feature representation of each branch's local decision. It clusters these decision features in recent failure and success memories to identify recurring local patterns. Using this pattern memory, PEARL selectively strengthens negative advantages at lower-reward decisions, while leaving positive advantages and the base policy optimizer unchanged. On AppWorld, PEARL improves over ARPO on Normal and Challenge tasks at both 4B and 8B scales. On ALFWorld, it achieves 97.9% mean success at 4B, 2.7 points above the strongest baseline, while remaining competitive at 8B. Ablations of shaping location, success-bank filtering, and memory window support the value of selective, recurring-pattern-based updates.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑