Ayanami0730/deep_research_bench

DeepResearch Bench: A Comprehensive Benchmark for Deep Research Agents

63
/ 100
Established

Comprises 100 PhD-level research tasks across 22 domains designed by domain experts, evaluated using multi-dimensional metrics including factuality, reasoning quality, and report comprehensiveness. Integrates with Hugging Face for dataset hosting and leaderboard management, with evaluation infrastructure powered by Gemini models for automated scoring across RACE (reasoning/analysis) and FACT (factuality) dimensions. Supports both proprietary and open-source deep research agents, enabling standardized comparison of agentic research capabilities against expert-curated reference standards.

637 stars. Actively maintained with 8 commits in the last 30 days.

No Package No Dependents
Maintenance 20 / 25
Adoption 10 / 25
Maturity 15 / 25
Community 18 / 25

How are scores calculated?

Stars

637

Forks

65

Language

Python

License

Apache-2.0

Last pushed

Mar 17, 2026

Commits (30d)

8

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/nlp/Ayanami0730/deep_research_bench"

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