RL-Index: Reinforcement Learning for Retrieval Index Reasoning
Summary
RL-Index proposes a reinforcement learning-based agentic indexing framework that shifts reasoning from query time to the indexing stage by augmenting documents with LLM-generated rationales, improving retrieval effectiveness and reducing online latency.
View Cached Full Text
Cached at: 06/25/26, 09:10 AM
Paper page - RL-Index: Reinforcement Learning for Retrieval Index Reasoning
Source: https://huggingface.co/papers/2606.16316
Abstract
RL-Index introduces an agentic indexing framework that shifts reasoning from query time to indexing stage by using LLM-generated rationales and reinforcement learning to improve retrieval effectiveness and reduce latency.
Retrieving external knowledge is essential for solving real-world tasks, yet it remains challenging when the relationship between a query and its relevant knowledge involves implicit and complex reasoning beyond surface-level semantic or lexical matching (e.g., mathematical problems relying on the same theorem or coding requiring deep reasoning). Existing approaches primarily rely onquery-side reasoning(e.g., query rewriting), which introduces significant online latency and underutilizes the opportunity to perform reasoning over the knowledge corpus itself (i.e.,index-side reasoning). In this paper, we propose RL-Index, an agentic indexing framework that formulatesretrieval index reasoningas areinforcement learningproblem. Instead of performing reasoning at query time, RL-Index shifts reasoning to the indexing stage by augmenting documents withLLM-generated rationalesthat explicitly encode the latent query-knowledge relationship. To optimize the quality of these rationales, we employGroup Relative Policy Optimization(GRPO) and useretrieval similarityas a verifiable reward signal, enabling direct optimization of indexing decisions for retrieval effectiveness. Extensive experiments on theBRIGHT benchmarkdemonstrate that RL-Index consistently improves both retrieval anddownstream question-answering performance, while significantly reducingonline inference latency. Moreover, the learned rationale augmentation generalizes across diverse retrievers and generators, highlighting its robustness as a plug-and-play indexing strategy across different retrieval systems.
View arXiv pageView PDFAdd to collection
Get this paper in your agent:
hf papers read 2606\.16316
Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash
Models citing this paper0
No model linking this paper
Cite arxiv.org/abs/2606.16316 in a model README.md to link it from this page.
Datasets citing this paper0
No dataset linking this paper
Cite arxiv.org/abs/2606.16316 in a dataset README.md to link it from this page.
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2606.16316 in a Space README.md to link it from this page.
Collections including this paper0
No Collection including this paper
Add this paper to acollectionto link it from this page.
Similar Articles
Rethinking Reasoning-Intensive Retrieval: Evaluating and Advancing Retrievers in Agentic Search Systems
The paper introduces BRIGHT-Pro, a new benchmark for reasoning-intensive retrieval, and RTriever-Synth, a synthetic corpus used to fine-tune RTriever-4B for improved performance in agentic search systems.
MCite-RL: Towards Reliable Multimodal RAG via Citation-enhanced Agentic Reinforcement Learning
MCite-RL is a citation-enhanced agentic reinforcement learning framework designed for reliable multimodal RAG, introducing iterative retrieval and reasoning for visual citations and a reward mechanism to jointly optimize answer accuracy and source traceability.
Retrieval Grounding Latent Reasoning for Dense Retrieval
Proposes Retrieval Grounding Latent Reasoning (RGLR), a latent reasoning framework for dense retrieval that explicitly connects intermediate latent transitions with retrieval improvements, outperforming baselines on reasoning-intensive tasks.
LatentRAG: Latent Reasoning and Retrieval for Efficient Agentic RAG
LatentRAG is a novel framework that shifts reasoning and retrieval for agentic RAG into continuous latent space, reducing inference latency by approximately 90% while maintaining performance comparable to explicit methods.
Rethinking the Necessity of Adaptive Retrieval-Augmented Generation through the Lens of Adaptive Listwise Ranking
This paper proposes AdaRankLLM, an adaptive retrieval framework that challenges the necessity of adaptive RAG by using listwise ranking to dynamically filter retrieved passages. The work shows that adaptive retrieval serves as a noise filter for weaker models while acting as a cost-efficiency optimizer for stronger models, with extensive experiments across multiple datasets and LLMs.