Summary

This repository serves as a comprehensive survey and curated collection of research papers on Reinforcement Learning (RL)-based agentic search systems. It targets researchers and practitioners in AI, LLMs, and information retrieval, providing a structured overview of how RL enables LLMs to intelligently plan, execute, and refine search queries for complex information-seeking tasks, thereby enhancing reasoning and evidence integration.

How It Works

The core approach involves treating information-seeking as a sequential decision-making process, where LLMs leverage RL to learn optimal strategies for when and how to search. This allows agents to dynamically issue, revise, and integrate search queries, learning from feedback to improve search efficiency, relevance, and reasoning capabilities. The repository categorizes and details various RL techniques, reward models (Outcome vs. Process), algorithms, and optimization scopes applied to agentic search systems.

Quick Start & Requirements

This repository is a curated collection of research papers and their summaries, not a runnable software project. It does not provide installation instructions or specific software requirements. It serves as a knowledge base for understanding RL-based agentic search systems. Links to individual research papers and some associated code repositories are provided within the tables.

Highlighted Details

• Features a comprehensive survey paper: "A Comprehensive Survey on Reinforcement Learning-based Agentic Search: Foundations, Roles, Optimizations, Evaluations, and Applications".
• Extensive tables detail representative research works, including their RL function roles, algorithms, reward types, optimization scopes (agent, step, module, system levels), datasets, and specific code links where available.
• Categorizes research by functional roles (e.g., Search Efficiency, Adaptive Search Decision, Reasoning-Search Interaction, Multi-Agent Collaboration, Tool and Knowledge Integration) and optimization scopes.
• Provides a detailed overview of evaluation metrics and benchmarks used in the field, including specific papers and their focuses.

Maintenance & Community

The repository states it is "actively maintaining this repository!". No specific community links (e.g., Discord, Slack) or contributor details are provided.

Licensing & Compatibility

No license information is specified in the provided README content. Compatibility for commercial use or closed-source linking is not addressed.

Limitations & Caveats

As a curated list of research papers, this repository does not offer a deployable software agent or framework. Its primary function is informational, providing an overview of academic research rather than a practical implementation. Users seeking to build or adopt such systems will need to refer to the individual papers and their respective codebases.