Benchmark
DualChem
Evaluate chemistry assistance and risk awareness when questions have both beneficial and harmful uses.
Explore the original evidence
Explore the original evaluation, its methods, and reported findings.
Open evalCoverage
600 expert-curated chemistry questions reviewed by PhD-level evaluators. Scale AI developed the benchmark with scope and taxonomy input from NIST CAISSI and U.S. government experts; Madhu Sehwag authored the announcement.
Methodology
Assess answer accuracy, assistance to chemistry graduates, recognition of dual-use risk, and agreement with expert severity judgments as separate measures. Creator-reported evaluation; no independent result verification or scores are included.