pat-jj/DeepRetrieval

[COLM’25] DeepRetrieval — 🔥 Training Search Agent by RLVR with Retrieval Outcome

51
/ 100
Established

Trains LLMs via reinforcement learning with retrieval metrics as rewards, eliminating the need for supervised query-augmentation pairs. The approach uses structured reasoning with explicit `` sections followed by final query formulation, enabling chain-of-thought reasoning before retrieval. Achieves 65% recall on publication search and supports diverse tasks including real search engines (PubMed, clinical trials), dense/sparse retrieval, and SQL database search with only 3B parameters.

696 stars.

No Package No Dependents
Maintenance 6 / 25
Adoption 10 / 25
Maturity 16 / 25
Community 19 / 25

How are scores calculated?

Stars

696

Forks

85

Language

Python

License

MIT

Category

math-reasoning-datasets

Last pushed

Oct 12, 2025

Commits (30d)

0

GitHub
Math Reasoning Datasets · 60 tools

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/llm-tools/pat-jj/DeepRetrieval"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.

Related tools

MMMU-Benchmark/MMMU

This repo contains evaluation code for the paper "MMMU: A Massive Multi-discipline Multimodal...

52

lupantech/MathVista

MathVista: data, code, and evaluation for Mathematical Reasoning in Visual Contexts

47

ise-uiuc/magicoder

[ICML'24] Magicoder: Empowering Code Generation with OSS-Instruct

45

x66ccff/liveideabench

[𝐍𝐚𝐭𝐮𝐫𝐞 𝐂𝐨𝐦𝐦𝐮𝐧𝐢𝐜𝐚𝐭𝐢𝐨𝐧𝐬] 🤖💡 LiveIdeaBench: Evaluating LLMs' Scientific Creativity and Idea...

42

IAAR-Shanghai/xVerify

xVerify: Efficient Answer Verifier for Reasoning Model Evaluations

40

Explore LLM Tools

All categories Trending LLM Tool directory Insights