SCIENCE UNDER AUTOMATION

AI science benchmarks and research evidence

Each record links the original result to the scientific requirements it informs.

• AlphaFold2 · 2021 paper · CASP14 (2020)
• GenCast · 2024 report
• AlphaDev · 2023
• AlphaProof + AlphaGeometry 2 · 2024
• A-Lab · 2023 paper, corrected January 2026
• Stokes et al. · Cell 2020
• BixBench3 · v2, September 2026
• CritPt · v4, May 2026
• FrontierScience · December 2025
• MULTICOM / CASP16
• Experiment-guided AlphaFold3
• AIFS Single 1.1.0
• ECMWF AIFS v2 operational release
• FrontierMath Tier 4 v2 / official CSV
• OpenAI mathematics release, updated October 15
• AlphaEvolve impact update
• AlphaEvolve technical report
• GeoNatureAgent v2
• LABBench2 v2
• GPT-5 / Ginkgo cell-free protein synthesis
• GPT-5 / Ginkgo technical report
• Multi-agent autonomous materials laboratories
• Rentosertib phase 2a trial
• CritPt independent leaderboard
• BixBench3 v2
• ExoMiner
• Coscientist
• MICrONS
• PhaseNet-DAS
• Prov-GigaPath
• MASAI
• SWE-Bench Pro Verified, v3
• Towards end-to-end automation of AI research
• METR task-completion time horizons
• An agentic artificially intelligent X-ray scientist
• Collider-Bench
• AutoLabs
• Autonomous mobile robots for exploratory synthetic chemistry
• Accelerating scientific discovery with Co-Scientist
• Autonomous enzyme engineering platform
• AMIE disease management
• Generative AI decision support in primary care
• SRT-H, v3
• Prospective multi-pathogen forecasting, v2
• Real-time respiratory outbreak early warning
• A foundation model for the Earth system
• GeoNatureAgent extended benchmark
• Physics-in-the-Loop
• CAD agents with finite-element feedback, v2
• SocSci-Repro-Bench
• Simulated preferences across 74 countries
• ReplicatorBench, v3
• Contextualizing ancient texts with generative neural networks
• AI-accelerated Nazca survey