HISR: Hindsight Information Modulated Segmental Process Rewards For Multi-turn Agentic Reinforcement Learning

TL;DR

HISR introduces a hindsight-based reward modulation technique for multi-turn reinforcement learning, improving credit assignment by aligning rewards with sub-goals and emphasizing significant segments, leading to better long-horizon decision-making.

Contribution

The paper presents a novel hindsight information approach to modulate segmental rewards, addressing reward sparsity and credit assignment issues in long-horizon RL tasks.

Findings

Improved performance on three benchmark tasks.

Enhanced credit assignment reliability.

Effective segmentation of long-horizon decision processes.

Abstract

While large language models excel in diverse domains, their performance on complex longhorizon agentic decision-making tasks remains limited. Most existing methods concentrate on designing effective reward models (RMs) to advance performance via multi-turn reinforcement learning. However, they suffer from delayed propagation in sparse outcome rewards and unreliable credit assignment with potentially overly fine-grained and unfocused turnlevel process rewards. In this paper, we propose (HISR) exploiting Hindsight Information to modulate Segmental process Rewards, which closely aligns rewards with sub-goals and underscores significant segments to enhance the reliability of credit assignment. Specifically, a segment-level process RM is presented to assign rewards for each sub-goal in the task, avoiding excessively granular allocation to turns. To emphasize significant segments in the trajectory, a hindsight model is devised to reflect the preference of performing a certain action after knowing the trajectory outcome. With this characteristic, we design the ratios of sequence likelihoods between hindsight and policy model to measure action importance. The ratios are subsequently employed to aggregate segment importance scores, which in turn modulate segmental process rewards, enhancing credit assignment reliability. Extensive experimental results on three publicly benchmarks demonstrate the validity of our method.