WildGuard

weights

open(allenai/wildguard on HF)

data

open(WildGuardMix corpus public)

code

partial(inference/serving package only

base

Mistral-7B

license

Apache-2.0(OSI)

Open weights under Apache-2.0 with the WildGuardMix training corpus published, which is more than most guardrail models release. The fine-tuning pipeline is not: the allenai/wildguard repo ships an inference package only - nine files, no trainer and no training config - the companion Safety-Eval repo is an evaluation suite, and the model card sends readers to the paper appendix for training details rather than to code. The rung above requires both released data and released code, so this scores 3; open data alone does not lift it.

• https://huggingface.co/allenai/wildguard recorded 2026-08-13

  Repo shows License - apache-2.0 and a dataset tag allenai/wildguardmix, gated auto (access granted on accepting the AI2 Responsible Use policy), with safetensors and pytorch_model shards downloadable. cardData declares datasets - allenai/wildguardmix, which is the published training corpus.

• https://api.github.com/repos/allenai/wildguard/contents recorded 2026-08-13

  Root listing of the project's own repo, which is the whole of what it ships as code. Nine entries (.gitignore, LICENSE.md, README.md, docs, examples, pyproject.toml, requirements.txt, tests, wildguard). No training directory, no trainer entrypoint, no SFT or DPO config. The `wildguard` package is the classifier wrapper and `examples/wildguard_filter` is a serving demo, so the published code is inference and serving only.

• https://github.com/allenai/wildguard/blob/main/README.md recorded 2026-08-13

  The project's own README installs with `pip install wildguard` and documents `load_wildguard()` plus `classify()`. Its only outbound links are the paper, the WildGuardMix dataset, the model, and the Safety-Eval companion repo, which it describes as holding "the details of evaluations run in the WildGuard paper". No training pipeline is offered or linked.

121,205 downloads in the trailing 30 days for the one declared artifact, allenai/wildguard, which bands at 100K-1M, level 3 on the software and model adoption scale.

• https://huggingface.co/api/models/allenai/wildguard recorded 2026-08-13

  121,205 downloads in the trailing 30 days for allenai/wildguard

Strong precision and low over-refusal; the joint refusal-detection head is a useful differentiator.

• https://huggingface.co/allenai/wildguard recorded 2026-08-13

  Card describes a one-stop moderation tool detecting prompt harm, response harm and model refusal across 13 risk subcategories, reported above GPT-4 on the paper's moderation benchmarks; tagged classifier, safety, moderation and arxiv 2406.18495.