Datasets:

jrosseruk
/

dare-domain-splits

Dataset card Files Files and versions

dare-domain-splits

87.6 GB

1 contributor

History: 115 commits

jrosseruk's picture

Add china_friendly eval for gemini_code

df5808b verified 6 months ago

activations
upload measurement activations: activations/gold/adapter/measurement_activations.parquet 6 months ago
ekfac
upload refusal EK-FAC scores (25K x 1K) 6 months ago
ekfac_contrastive
upload contrastive EK-FAC scores for valid_feelings 6 months ago
ekfac_standard
upload standard EK-FAC scores for valid_feelings (top 200 validate) 6 months ago
model_comparison
Add china_friendly eval for gemini_code 6 months ago
refusal_dataset
Upload refusal_dataset/refusal_indexed.jsonl with huggingface_hub 6 months ago
retrain
retrain probe/ethical_frameworks pct50 6 months ago
retrain_coding_plus_refusal
coding + refusal top 25% (llm_judge): adapter + evals 6 months ago
retrain_refusal
retrain refusal contrast_diff pct25 6 months ago
retrain_sft_minus_wildchat
SFT minus wildchat: adapter + evals 6 months ago
retrain_sft_minus_wildchat_coconot
SFT minus wildchat+coconot retrain + evals 6 months ago
retrain_validate_feelings
retrain valid_feelings ekfac_contrastive pct25 6 months ago
steer
steering results: 5 behaviors x 5 alphas x 25 queries (think prefill, hypothesis queries) 6 months ago
train
Add both_sides + liberal_lean evals 6 months ago
validate_feelings
Upload validate_feelings/validate_feelings_indexed.jsonl with huggingface_hub 6 months ago
.gitattributes

31.3 kB
Add china_friendly eval for gemini_code 6 months ago