PRAISE: Prefix-Based Rollout Reuse in Agentic Search Training
This addresses data efficiency and credit assignment issues in agentic search training for complex tasks like multi-hop QA, representing an incremental advancement.
The paper tackled the problem of inefficient training and reward sparsity in agentic search for multi-hop question answering by proposing PRAISE, which reuses search trajectory prefixes to generate additional training data and intermediate rewards, resulting in consistent performance improvements on benchmarks.
In agentic search, large language models (LLMs) are trained to perform multi-turn retrieval and reasoning for complex tasks such as multi-hop question answering (QA). However, current search-based Reinforcement Learning (RL) methods suffer from two core limitations: expensive long-horizon rollouts are under-utilized during training, and supervision is typically available only at the final answer, resulting in severe reward sparsity. We present Prefix-based Rollout reuse for Agentic search with Intermediate Step rEwards (PRAISE), a framework for improving both data efficiency and credit assignment in agentic search training. Given a complete search trajectory, PRAISE extracts prefix states at different search turns, elicits intermediate answers from them, and uses these prefixes both to construct additional training trajectories and to derive step-level rewards from performance differences across prefixes. Our method uses a single shared model for both search policy learning and prefix answer evaluation, enabling joint optimization without extra human annotations or a separate reward model. Experiments on multi-hop QA benchmarks show that PRAISE consistently improves performance over strong baselines.