GAMOW LABS / EVALUATIONS
Benchmarks
AI agents, tested on real scientific problems.
From rare disease diagnosis to planning and running experiments.
Independent evals are the best disinfectant
01 / RARE DISEASE DIAGNOSIS
RareBench
Can an agent identify the causal findings in a rare disease case?
Complete causal recall
Loading graded results…
All required causal findings within the first K ranked findings; K is the number of required findings for each case.
- Loading results…
Whiskers show 95% Wilson intervals. Invalid outputs count as failures.
02 / EXPERIMENTAL REASONING
LabBench
Can an agent reason from lab evidence to the right next experiment?
Mean task score
Loading graded results…
The average fraction of rubric criteria passed per task. Each task has equal weight.
- Loading results…
Only complete, fully reviewed 20-task configurations are shown.
03 / ROBOTIC LABORATORY WORKFLOWS
LabRobotBench
Can an agent operate a robotic laboratory and recover interrupted experiments?
Mean score across our held-out subset
Research preview · V0.1
- GPT 6.1 SolCodex46.7%
- Opus 5.5Claude Code23.0%
- Gemini 4 ArgonGemini CLI20.9%