FractalAIResearchLabs/Fathom-DeepResearch

Fathom-DeepResearch: Unlocking Long Horizon Information Retrieval And Synthesis For SLMs

40
/ 100
Emerging

Implements a two-model agentic pipeline: **Fathom-Search-4B** performs iterative web retrieval and evidence verification through live search APIs (Serper, Jina, Crawl4AI), while **Fathom-Synthesizer-4B** converts search traces into citation-dense reports via a plan-then-write protocol. Training leverages RAPO (Reward-Aware Policy Optimization) with step-level steerable rewards to stabilize multi-turn RL, plus a multi-agent self-play framework (DuetQA dataset) that generates search-enforcing QA pairs impossible to answer without live web access.

No commits in the last 6 months.

Stale 6m No Package No Dependents
Maintenance 2 / 25
Adoption 8 / 25
Maturity 15 / 25
Community 15 / 25

How are scores calculated?

Stars

55

Forks

9

Language

Python

License

MIT

Last pushed

Oct 07, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/agents/FractalAIResearchLabs/Fathom-DeepResearch"

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