# Rethinking On-policy Optimization for Query Augmentation

arXiv: https://arxiv.org/abs/2510.17139 · HTML: https://mlsanity.com/paper/2510.17139/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Knowledge- and tool-augmented reasoning, one of 17 directions in LLM reasoning.
- Builds on 30 of its 94 references that are in the map.
- First posted on arXiv 20 Oct 2025.

## Placed in

- LLM reasoning / [Knowledge- and tool-augmented reasoning](https://mlsanity.com/lr/knowledge-and-tool-augmented-reasoning/)

## Builds on (most cited first, 30 of 30)

- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://mlsanity.com/paper/2501.12948/) (2025)
- [Qwen2 Technical Report](https://mlsanity.com/paper/2407.10671/) (2024)
- [MS MARCO: A Human Generated MAchine Reading COmprehension Dataset](https://mlsanity.com/paper/1611.09268/) (2016)
- [Search-R1: Training LLMs to Reason and Leverage Search Engines with Reinforcement Learning](https://mlsanity.com/paper/2503.09516/) (2025)
- [gpt-oss-120b & gpt-oss-20b Model Card](https://mlsanity.com/paper/2508.10925/) (2025)
- [Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Models](https://mlsanity.com/paper/2506.05176/) (2025)
- [Text Embeddings by Weakly-Supervised Contrastive Pre-training](https://mlsanity.com/paper/2212.03533/) (2022)
- [Kimi k1.5: Scaling Reinforcement Learning with LLMs](https://mlsanity.com/paper/2501.12599/) (2025)
- [Tulu 3: Pushing Frontiers in Open Language Model Post-Training](https://mlsanity.com/paper/2411.15124/) (2024)
- [Rubrics as Rewards: Reinforcement Learning Beyond Verifiable Domains](https://mlsanity.com/paper/2507.17746/) (2025)
- [R1-Searcher: Incentivizing the Search Capability in LLMs via Reinforcement Learning](https://mlsanity.com/paper/2503.05592/) (2025)
- [Spurious Rewards: Rethinking Training Signals in RLVR](https://mlsanity.com/paper/2506.10947/) (2025)
- [BrowseComp-Plus: A More Fair and Transparent Evaluation Benchmark of Deep-Research Agent](https://mlsanity.com/paper/2508.06600/) (2025)
- [ReSearch: Learning to Reason with Search for LLMs via Reinforcement Learning](https://mlsanity.com/paper/2503.19470/) (2025)
- [DeepRetrieval: Hacking Real Search Engines and Retrievers with Large Language Models via Reinforcement Learning](https://mlsanity.com/paper/2503.00223/) (2025)
- [Pyserini: An Easy-to-Use Python Toolkit to Support Replicable IR Research with Sparse and Dense Representations](https://mlsanity.com/paper/2102.10073/) (2021)
- [A Replication Study of Dense Passage Retriever](https://mlsanity.com/paper/2104.05740/) (2021)
- [LLM-based relevance assessment still can't replace human relevance assessment](https://mlsanity.com/paper/2412.17156/) (2024)
- [ParallelSearch: Train your LLMs to Decompose Query and Search Sub-queries in Parallel with Reinforcement Learning](https://mlsanity.com/paper/2508.09303/) (2025)
- [Overview of the TREC 2022 deep learning track](https://mlsanity.com/paper/2507.10865/) (2025)
- [A Comprehensive Survey on Reinforcement Learning-based Agentic Search: Foundations, Roles, Optimizations, Evaluations, and Applications](https://mlsanity.com/paper/2510.16724/) (2025)
- [RankMamba: Benchmarking Mamba's Document Ranking Performance in the Era of Transformers](https://mlsanity.com/paper/2403.18276/) (2024)
- [Deep Reinforced Query Reformulation for Information Retrieval](https://mlsanity.com/paper/2007.07987/) (2020)
- [Reinforcement Learning to Rank Using Coarse-grained Rewards](https://mlsanity.com/paper/2208.07563/) (2022)
- [Query Expansion in the Age of Pre-trained and Large Language Models: A Comprehensive Survey](https://mlsanity.com/paper/2509.07794/) (2025)
- [Beyond the limitation of a single query: Train your LLM for query expansion with Reinforcement Learning](https://mlsanity.com/paper/2510.10009/) (2025)
- [Hybrid Deep Searcher: Scalable Parallel and Sequential Search Reasoning](https://mlsanity.com/paper/2508.19113/) (2025)

## Built on by (most cited first, 3 of 3)

- [CLEAR: Context Augmentation from Contrastive Learning of Experience via Agentic Reflection](https://mlsanity.com/paper/2604.07487/) (2026)
- [Beyond Supervised Clarification: Input Rewriting with LLMs for Dialogue Discourse Parsing](https://mlsanity.com/paper/2607.01964/) (2026)
- [Tevatron-Elastic: A Unified Abstraction for Training Elastic Retrievers and Rerankers](https://mlsanity.com/paper/2608.08809/) (2026)

## Abstract (arXiv)

Recent advances in large language models (LLMs) have led to a surge of interest in query augmentation for information retrieval (IR). Two main approaches have emerged. The first prompts LLMs to generate answers or pseudo-documents that serve as new queries, relying purely on the model's parametric knowledge or contextual information. The second applies reinforcement learning (RL) to fine-tune LLMs for query rewriting, directly optimizing retrieval metrics. While having respective advantages and limitations, the two approaches have not been compared under consistent experimental conditions. In this work, we present the first systematic comparison of prompting-based and RL-based query augmentation across diverse benchmarks, including evidence-seeking, ad hoc, and tool retrieval. Our key finding is that under a compute-aware comparison setting, simple, training-free query augmentation often performs on par with, or even surpasses, more expensive RL-based counterparts, especially when using powerful LLMs. Motivated by this discovery, we introduce a novel hybrid method, On-policy Pseudo-document Query Expansion (OPQE), in which the LLM policy learns to generate a pseudo-document that maximizes retrieval performance, rather than rewriting the query, thus merging the flexibility and generative structure of prompting with the targeted optimization of RL. We show OPQE outperforms both standalone prompting and RL-based rewriting, demonstrating that a synergistic approach yields the best results. We open source our implementation to facilitate reproducibility.