Termsslug: experimental-controls

Experimental Controls

实验对照

Reference groups and calibration conditions used to rule out confounds and validate experimental results.

Curated Intelligence

1 items · Page 1/1
OpenAI 官方博客T1·Agents
58 pts
Anthropic and Ligo Release BioBasicsBench to Test AI Agents on Wet-Lab Analysis

Anthropic and enzyme-design firm Ligo introduced BioBasicsBench, a 100-task benchmark measuring whether AI agents can interpret raw wet-lab data, spot experimental errors, and recognize when evidence fails to support a conclusion. It targets the messy reality of failed runs, ambiguous images and conflicting observations that paper-derived benchmarks tend to miss.