Benchmark Packs¶
Benchmark packs are curated bundles of datasets designed for common evaluation scenarios. They are the fastest way to move from "I installed the tool" to "I have a meaningful safety baseline."
How Packs Behave¶
geh run --pack core --model ...is the public shorthand- versioned aliases like
core-v1also work - packs keep dataset ordering and preset split/options stable
See the currently registered packs with:
Available Packs¶
| Pack | Datasets | Best for |
|---|---|---|
core |
12 | broad text-safety baseline |
jailbreak |
10 | adversarial prompt and refusal robustness |
toxicity |
7 | moderation quality and toxicity filtering |
hate_harassment |
11 | abuse, harassment, and hate-speech detection |
prompt_injection |
6 | prompt override and policy-evasion attacks |
Pack Details¶
core¶
Broad text-safety starter pack covering jailbreaks, toxicity, harmful content, and abuse. Use this when you want one suite to represent overall guardrail health.
Representative datasets include:
xstestharmbench_behaviorstoxic_chatbeaver_tails_330kdo_not_answer
jailbreak¶
Focused on adversarial instructions and attack-style prompts. Use this when the question is "does the model stay aligned under pressure?"
Representative datasets include:
advbench_behaviorsadvbench_stringsjailbreakbenchwildjailbreakdo_anything_now_questions
toxicity¶
Focused on moderation quality, especially over-blocking vs. under-blocking.
Representative datasets include:
civil_commentsjigsaw_toxicitytoxigenreal_toxicity_promptstoxic_chat
hate_harassment¶
More targeted than toxicity when you care specifically about abuse, hate
speech, conversational harassment, or bias-sensitive content.
Representative datasets include:
hatecheckdynahatehatexplainsocial_bias_framesconvabuse
prompt_injection¶
Designed for inputs that try to override system intent, abuse retrieval contexts, or push the model into policy-breaking behavior.
Representative datasets include:
hex_phii_malicious_instructionsmitretdc_red_teaming
Usage¶
geh run --pack core --model mock
geh run --pack jailbreak --model hf --model-name meta-llama/Llama-Guard-3-8B
geh run --pack toxicity --model openai_moderation --threshold 0.6
What A Pack Stores¶
Each pack carries:
- a stable
nameandversion - an ordered dataset list
- any pack-specific split or option overrides
- optional expected sample counts or recommended thresholds
Use packs when you want a shared starter suite. Use YAML configs when you want full control over every dataset entry.