Today for AI
HOT RADAR
llmHEAT 10.8°

AI CURATED HOT STORY · 11/14/2026

Stanford-led Team Launches SPU Benchmark with 4,207 Wet-Lab Tests

1 reports archived1 independent sourcesupdated 11/14/2026, 17:58:22
Synthesis & Latest Updates
1 Sources Cross-Validated

UniPat AI, together with Stanford, Harvard, Cornell and over ten other institutions, released Science Prediction Universe (SPU), a benchmark of 116 tasks and 4,207 real wet-lab experiments that tests whether AI can predict experimental outcomes before they are run. Models were given an H20 GPU and an 8-hour workday to autonomously use retrieval and compute resources, yet even Opus 5.5 and GPT-6 Astra remained far from usable predictive accuracy, and a 5x compute increase did not help. The benchmark turns hard-to-quantify scientific judgment into a measurable, comparable score.

LATEST/UniPat AI, together with Stanford, Harvard, Cornell and over ten other institutions, released Science Prediction Universe (SPU), a benchmark of 116 tasks and 4,207 real wet-lab experiments that tests whether AI can predict experimental outcomes before they are run. Models were given an H20 GPU and an 8-hour workday to autonomously use retrieval and compute resources, yet even Opus 5.5 and GPT-6 Astra remained far from usable predictive accuracy, and a 5x compute increase did not help. The benchmark turns hard-to-quantify scientific judgment into a measurable, comparable score.

Hourly heat curve

6 fully observed hours
Now
10.8
Peak
12.018:00
24h change
–

The line compares only sources observed throughout; gap hours are interpolated to keep the trend continuous.

Coverage timeline

Total 1 reports · Latest first
  1. 量子位官网T2·78 pts

    Stanford-led Team Launches SPU Benchmark with 4,207 Wet-Lab Tests

    • SPU spans 116 tasks and 4,207 wet-lab experiments across chemistry, materials, biology and medicine, with advisors and lab partners from Stanford, Harvard and Cornell
    • Models working autonomously for 8 hours on a single H20 GPU could not reliably predict experimental outcomes
    • A 5x compute increase brought no substantive gains, indicating the bottleneck is scientific judgment rather than compute