Pith. sign in

REVIEW 15 cited by

DeepRetrieval: Hacking Real Search Engines and Retrievers with Large Language Models via Reinforcement Learning

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2503.00223 v3 pith:P37DUGNA submitted 2025-02-28 cs.IR

Pengcheng Jiang , Jiacheng Lin , Lang Cao , Runchu Tian , SeongKu Kang , Zifeng Wang , Jimeng Sun , Jiawei Han This is my paper
classification cs.IR
keywords retrievalsearchdeepretrievaldatainformationlargelearningmodels
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Information retrieval systems are crucial for enabling effective access to large document collections. Recent approaches have leveraged Large Language Models (LLMs) to enhance retrieval performance through query augmentation, but often rely on expensive supervised learning or distillation techniques that require significant computational resources and hand-labeled data. We introduce DeepRetrieval, a reinforcement learning (RL) approach that trains LLMs for query generation through trial and error without supervised data (reference query). Using retrieval metrics as rewards, our system generates queries that maximize retrieval performance. DeepRetrieval outperforms leading methods on literature search with 65.07% (vs. previous SOTA 24.68%) recall for publication search and 63.18% (vs. previous SOTA 32.11%) recall for trial search using real-world search engines. DeepRetrieval also dominates in evidence-seeking retrieval, classic information retrieval and SQL database search. With only 3B parameters, it outperforms industry-leading models like GPT-4o and Claude-3.5-Sonnet on 11/13 datasets. These results demonstrate that our RL approach offers a more efficient and effective paradigm for information retrieval. Our data and code are available at: https://github.com/pat-jj/DeepRetrieval.

Discussion (0). Sign in to comment.

Forward citations

Cited by 15 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Contextual Information Policy Optimization for Search Agents

    cs.AI 2026-08 conditional novelty 7.0 of 10

    CIPO trains search agents with a dense reward that measures how much a reasoning step depends on newly retrieved evidence, improving F1 across seven QA benchmarks.

  2. Antares: Foundation Models for Agentic Vulnerability Localization

    cs.CR 2026-08 conditional novelty 6.0 of 10

    Antares-3B, a 3B model trained with SFT plus GRPO, matches GPT-5.5 on repository-scale vulnerability localization at roughly 1/100th the inference cost.

  3. RICE-PO: Turning Retrieval Interactions into Credit Signals for Reasoning Agents

    cs.CL 2026-05 unverdicted novelty 6.0 of 10

    RICE-PO trains retrieval agents using retrieval scores of intermediate summaries as local rewards, gated by influence and residual-stability estimates, outperforming group-based RL baselines on BRIGHT and BEIR.

  4. Revolutionizing Reinforcement Learning Framework for Diffusion Large Language Models

    cs.CL 2025-09 conditional novelty 6.0 of 10

    TraceRL aligns the RL objective with the DLM's actual step-by-step decoding, producing TraDo-4B/8B models that beat autoregressive baselines on math reasoning.

  5. VerifyBench: A Systematic Benchmark for Evaluating Reasoning Verifiers Across Domains

    cs.AI 2025-07 conditional novelty 6.0 of 10

    VerifyBench systematically compares specialized and general LLM verifiers across 3,989 math, physics, chemistry, and biology questions, finding an accuracy-recall tradeoff and strong sensitivity to input structure.

  6. Harnessing the Power of Reinforcement Learning for Language-Model-Based Information Retriever via Query-Document Co-Augmentation

    cs.IR 2025-06 conditional novelty 6.0 of 10

    A bidirectional RL framework jointly optimizes LLM query and document augmentation and improves retrieval accuracy on BEIR benchmarks beyond query-only rewriting.

  7. TongSearch-QR: Reinforced Query Reasoning for Retrieval

    cs.IR 2025-06 conditional novelty 6.0 of 10

    TongSearch-QR trains 1.5B and 7B models with GRPO and a frozen-embedding reward to rewrite queries, reaching 27.9 nDCG@10 on BRIGHT with BM25, above GPT-4o's 26.5.

  8. Cell-o1: Training LLMs to Solve Single-Cell Reasoning Puzzles with Reinforcement Learning

    cs.CL 2025-06 conditional novelty 6.0 of 10

    A 7B model trained with reasoning distillation and reinforcement learning reaches 32.9% batch-level accuracy on a new single-cell annotation benchmark, versus 19.0% for OpenAI's o1.

  9. Training LLMs for EHR-Based Reasoning Tasks via Reinforcement Learning

    cs.CL 2025-05 conditional novelty 6.0 of 10

    A 3B model trained with a small SFT warm-up followed by verifiable-reward RL matches or exceeds far larger models on EHR-based medical calculation, trial matching, and diagnosis tasks.

  10. SLMs as Multi-Agent Routers: A Progressive SFT and Reinforcement Learning Approach

    cs.CL 2026-07 reject novelty 5.0 of 10

    A 0.6B router trained by SFT+RL on retrieval-quality rewards reaches 0.771 NDCG@10 across 11 agents, beating intent-prompted LLMs and cutting latency by 82%.

  11. Rethinking On-policy Optimization for Query Augmentation

    cs.CL 2025-10 conditional novelty 5.0 of 10

    Simple zero-shot prompt-based query expansion matches costly RL-trained query rewriting in retrieval benchmarks, and training the RL policy to generate pseudo-documents (OPQE) yields the best overall scores.

  12. Domain-Aware RAG: MoL-Enhanced RL for Efficient Training and Scalable Retrieval

    cs.CL 2025-09 conditional novelty 5.0 of 10

    A two-stage RAG training pipeline, MoL continual pre-training plus GRPO reinforcement learning with single-passage training and multi-passage inference, reports state-of-the-art retrieval recall on NFCORPUS and SCIFAC...

  13. Beyond Independent Passages: Adaptive Passage Combination Retrieval for Retrieval Augmented Open-Domain Question Answering

    cs.CL 2025-07 conditional novelty 5.0 of 10

    AdaPCR jointly retrieves and reranks passage pairs for open-domain QA, showing small EM/F1 gains over an in-context retrieval baseline, mostly on multi-hop HotpotQA.

  14. VRAG-RL: Empower Vision-Perception-Based RAG for Visually Rich Information Understanding via Iterative Reasoning with Reinforcement Learning

    cs.CL 2025-05 conditional novelty 5.0 of 10

    VRAG-RL uses GRPO reinforcement learning with visual cropping actions and a retrieval-aware reward to improve vision-language RAG agents on document benchmarks.

  15. Deep Research Agents: A Systematic Examination And Roadmap

    cs.AI 2025-06 conditional novelty 4.0 of 10

    A survey that organizes LLM-powered deep research agents into static versus dynamic workflows and single versus multi agent architectures, and reviews their benchmarks and open challenges.

Pith tools