Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Freshness-Aware Prioritized Experience Replay for LLM/VLM Reinforcement Learning

About

Reinforcement Learning (RL) has achieved impressive success in post-training Large Language Models (LLMs) and Vision-Language Models (VLMs), with on-policy algorithms such as PPO, GRPO, and REINFORCE++ serving as the dominant paradigm. However, these methods discard all collected trajectories after a single gradient update, resulting in poor sample efficiency, particularly wasteful for agentic tasks where multi-turn environment interactions are expensive. While Experience Replay drives sample efficiency in classic RL by allowing agents to reuse past trajectories and prioritize informative ones, directly applying Prioritized Experience Replay (PER) to LLMs fails. The rapid policy evolution of billion-parameter models renders stored priorities stale, causing old high-priority trajectories to dominate sampling long after they have become uninformative. We propose Freshness-Aware PER, which addresses this priority staleness problem by augmenting any PER-based priority with a multiplicative exponential age decay grounded in effective sample size analysis. To the best of our knowledge, Freshness-Aware PER is the first work to successfully apply PER to LLM/VLM reinforcement learning. We evaluate on eight multi-step agentic, reasoning, and math competition tasks with 0.5B, 3B, and 7B models. Freshness-Aware PER significantly outperforms on-policy baselines, achieving +46% on NQ Search, +367% on Sokoban, and +133% on VLM FrozenLake, while standard PER without age decay consistently degrades performance. Our code is publicly available at https://github.com/Vision-CAIR/Freshness-Aware-PER.

Weiyu Ma, Yongcheng Zeng, Yan Song, Xinyu Cui, Jian Zhao, Xuhui Liu, Mohamed Elhoseiny• 2026

Related benchmarks

TaskDatasetResultRank
Question AnsweringNQ
Exact Match74.2
101
Geometric Question AnsweringGeoQA
Success Rate48.1
3
Mathematical Problem SolvingAIME
AIME Success Rate24.2
3
NavigationFrozenLake LLM
Success Rate30.5
3
NavigationFrozenLake VLM
Success Rate63
3
Puzzle SolvingSokoban Simple
Score2.304
3
Puzzle SolvingSokoban Hard
Score0.512
3
Showing 7 of 7 rows

Other info

Follow for update