Han and colleagues: WildGuard moderation and refusal classification

What changed

Developed a multi-task moderation model and associated training/evaluation resources that distinguish prompt harmfulness, response harmfulness and whether a response refuses the request, including adversarially phrased inputs.

Scientific scope

Developed a multi-task moderation model and associated training/evaluation resources that distinguish prompt harmfulness, response harmfulness and whether a response refuses the request, including adversarially phrased inputs. A refusal can be appropriate or excessive and is not itself a harmfulness label. Reported comparisons are benchmark- and model-specific; synthetic data and human judgments remain fallible.

Evidence

The linked original works support the scoped statement and contribution attribution. WRS has not independently replayed experiments or reconstructed a full mathematical proof. A refusal can be appropriate or excessive and is not itself a harmfulness label. Reported comparisons are benchmark- and model-specific; synthetic data and human judgments remain fallible.

Contributors

• The named authors in the research record · SCIENCE: Source-listed authors for the exact contribution; experimental subjects are not autonomous scientific authors. Source support and predecessor work retain separate roles. Individual effort shares and unresolved identities are not inferred.

• WildGuard: Open One-Stop Moderation Tools for Safety Risks, Jailbreaks, and Refusals of LLMs

Provisional WRS assessment, edition 2026-10-23.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction