NeFut Logo NeFut
Admin Login

[CS.AI] From Outcomes to Actions: Enhancing Long-Horizon Language Agent Training

Published at: 2026-07-22 22:00 Last updated: 2026-07-23 12:33
#AI #Machine Learning #optimization

Reinforcement learning (RL) has become a widely adopted technique for improving large language models (LLMs) on complex tasks. However, existing RL methods face challenges in training agents with longer-horizon interactions. A major bottleneck is distinguishing the contribution of different actions in long-horizon interactions, leading to high optimization variance.

To address this, we introduce a novel policy gradient method, Hindsight Policy Optimization (HPO), which projects both the current policy distribution and the hindsight distribution into an intent space and extracts low-variance learning signals from the Wasserstein distance between them.

We theoretically and empirically demonstrate that aggregating semantically similar states and actions in the intent space yields a bounded-variance estimator and improves policy performance stably.

Our code is available online.

Original Source: https://arxiv.org/abs/2607.16257

[h] Back to Home