HOT RADAR
agentHEAT 11.4°

AI CURATED HOT STORY · 11/21/2026

Google and Johns Hopkins Unveil Pro2 Agent That Tracks Bottlenecks Before Coding

1 reports archived1 independent sourcesupdated 11/21/2026, 03:49:29
Synthesis & Latest Updates
1 Sources Cross-Validated

Google and Johns Hopkins researchers introduce Pro2 Agent, which has ML research agents first build a 'Problem State' capturing suspected bottlenecks, candidate fixes and confidence levels, then run single-variable experiments to update beliefs before committing code changes. It reaches a 76.0% medal rate across all 75 MLE-Bench tasks within 48 hours, the highest reported, and medals on 57.6% of MLE-Bench Lite tasks with its first solution versus 28.8% for MARS.

LATEST/Google and Johns Hopkins researchers introduce Pro2 Agent, which has ML research agents first build a 'Problem State' capturing suspected bottlenecks, candidate fixes and confidence levels, then run single-variable experiments to update beliefs before committing code changes. It reaches a 76.0% medal rate across all 75 MLE-Bench tasks within 48 hours, the highest reported, and medals on 57.6% of MLE-Bench Lite tasks with its first solution versus 28.8% for MARS.

Hourly heat curve

13 fully observed hours
Now
11.4
Peak
15.005:00
24h change
-24%

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

Coverage timeline

Total 1 reports · Latest first
  1. Rohan Paul (@rohanpaul_ai)T2·78 pts
    • Pro2 Agent explicitly tracks bottlenecks, candidate fixes and confidence in a 'Problem State' to avoid tweaking fundamentally wrong solutions
    • Each experiment changes a single variable, and results update beliefs before any commit
    • 76.0% medal rate on all 75 MLE-Bench tasks within 48 hours; 57.6% first-solution medal rate on MLE-Bench Lite vs 28.8% for MARS