The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearning
Hazardous knowledge proxies and unlearning evaluation.
Verify byline in the original ↗PUBLICATION BYLINE
Indexed publications with this author name. Affiliation and author identity require publication-specific evidence.
Hazardous knowledge proxies and unlearning evaluation.
Verify byline in the original ↗Robust refusal under automated red teaming.
Verify byline in the original ↗Separates honesty from factual accuracy using repeated, human-collected test situations.
Verify byline in the original ↗