ASI-Bench

What is ASI-Bench?

ASI-Bench evaluates AI systems on 60 project-level scientific research tasks across 11 domains, with four guidance levels B1-B4 measuring autonomous execution…

Released

2026-08-18

Evaluates

AI Scientist, Agents, Agents & Tool Use, Autonomous execution, Reasoning, Research planning, Scientific Research & AI for Science, Scientific discovery

Openness

unknown

Importer

Claire Radar

Review status

source-reviewed

Reported scores

0

Source provenance

• Original evidence https://arxiv.org/abs/2608.17271

ASI-Bench paper, code and dataset

• Paper https://arxiv.org/abs/2608.17271
• Code https://github.com/apexin-ai/ASI-Bench
• Dataset https://huggingface.co/datasets/Apexintelligence-AI/ASI-Bench-seed31415
• Website https://asibench.apexin.ai

Which sources cite ASI-Bench?

• ASI-Bench

Related AI Scientist, Agents, Agents & Tool Use, Autonomous execution, Reasoning, Research planning, Scientific Research & AI for Science, Scientific discovery benchmarks

• $\tau^\tau$-Bench: An Environment for End-To-End, Realistic Agent Construction
• A-OKVQA
• A3S-Bench
• ActiveFly-Bench
• ActiveSciBench
• AdaPlanBench
• Adaptive Adversaries
• ADeptS-Bench