{"adoption": {"forks": 322, "observed_at": "2026-08-28T04:06:19.249893+00:00", "stars": 2149}, "canonical_url": "https://ross.abutalabs.com/products/remove-refusals-with-transformers", "card": {"archived": false, "artifact_type": "library", "description": "Implements harmful/harmless refusal removal using pure HF Transformers", "domain": ["large-language-models", "machine-learning", "artificial-intelligence"], "enriched": true, "function": ["llm-inference", "machine-learning", "deep-learning"], "health_score": 53, "homepage": null, "language": "Python", "license": "Apache-2.0", "license_family": "permissive", "maturity": "experimental", "member_repos": ["Sumandora/remove-refusals-with-transformers"], "name": "Sumandora/remove-refusals-with-transformers", "platform": ["python", "windows"], "pushed_at": "2025-11-27T10:27:12+00:00", "repo": "Sumandora/remove-refusals-with-transformers", "stars": 2149, "tags": ["refusal-direction", "transformerlens-alternative", "model-ablation", "safety-research", "proof-of-concept", "huggingface-transformers", "gpu", "linux", "macos"], "topics": [], "urls": [], "use_cases": ["remove refusals from an LLM without TransformerLens", "ablate the refusal direction in any HF Transformers model", "study how refusal is mediated by a single direction in LLMs", "test refusal removal on small models with limited GPU memory", "compare refusal behavior before and after direction ablation"], "what_it_is": "A proof-of-concept Python script that removes refusal behavior from LLMs by ablating the refusal direction using pure Hugging Face Transformers, without requiring TransformerLens. It computes the refusal direction from harmful/harmless instruction datasets and applies it during inference.", "when_to_avoid": ["you need a production-ready or maintained tool rather than a proof of concept", "your model uses a custom layer structure like some Qwen implementations", "you need broad model compatibility guarantees or extensive documentation"], "when_to_choose": ["you want to experiment with refusal-direction ablation using only Hugging Face Transformers", "your model isn't supported by TransformerLens", "you're doing interpretability or safety research on consumer GPUs with small models"]}, "data_as_of": "2026-08-30T08:39:29.467469+00:00", "members": [{"path": "/products/remove-refusals-with-transformers", "repo": "Sumandora/remove-refusals-with-transformers", "role": "main", "score": 48}], "provenance": {"archived": {"kind": "observed", "observed_at": "2026-08-28T04:06:19.249893+00:00", "source": "github"}, "artifact_type": {"confidence": null, "enriched_at": "2026-08-30T02:50:59.155304+00:00", "kind": "inferred", "prompt_version": 1, "sources": [{"content_hash": "ca7720eea288291430efeea567081fb24397bfa971a6140e1840742d2862cb9d", "fetched_at": "2026-08-28T04:06:19.249893+00:00", "kind": "readme", "missing": false, "url": "https://github.com/Sumandora/remove-refusals-with-transformers"}], "taxonomy_version": 1}, "description": {"kind": "observed", "observed_at": "2026-08-28T04:06:19.249893+00:00", "source": "github"}, "domain": {"confidence": null, "enriched_at": "2026-08-30T02:50:59.155304+00:00", "kind": "inferred", "prompt_version": 1, "sources": [{"content_hash": "ca7720eea288291430efeea567081fb24397bfa971a6140e1840742d2862cb9d", "fetched_at": "2026-08-28T04:06:19.249893+00:00", "kind": "readme", "missing": false, "url": "https://github.com/Sumandora/remove-refusals-with-transformers"}], "taxonomy_version": 1}, "enriched": {"inputs": [], "kind": "computed", "method": "enrichment_status"}, "function": {"confidence": null, "enriched_at": "2026-08-30T02:50:59.155304+00:00", "kind": "inferred", "prompt_version": 1, "sources": [{"content_hash": "ca7720eea288291430efeea567081fb24397bfa971a6140e1840742d2862cb9d", "fetched_at": "2026-08-28T04:06:19.249893+00:00", "kind": "readme", "missing": false, "url": "https://github.com/Sumandora/remove-refusals-with-transformers"}], "taxonomy_version": 1}, "health_score": {"inputs": ["days_since_push", "days_since_release", "archived"], "kind": "computed", "method": "health_v1"}, "homepage": {"kind": "observed", "observed_at": "2026-08-28T04:06:19.249893+00:00", "source": "github"}, "language": {"kind": "observed", "observed_at": "2026-08-28T04:06:19.249893+00:00", "source": "github"}, "license": {"kind": "observed", "observed_at": "2026-08-28T04:06:19.249893+00:00", "source": "github"}, "license_family": {"inputs": ["license"], "kind": "computed", "method": "license_family"}, "maturity": {"confidence": null, "enriched_at": "2026-08-30T02:50:59.155304+00:00", "kind": "inferred", "prompt_version": 1, "sources": [{"content_hash": "ca7720eea288291430efeea567081fb24397bfa971a6140e1840742d2862cb9d", "fetched_at": "2026-08-28T04:06:19.249893+00:00", "kind": "readme", "missing": false, "url": "https://github.com/Sumandora/remove-refusals-with-transformers"}], "taxonomy_version": 1}, "member_repos": {"kind": "observed", "observed_at": "2026-08-28T04:06:19.249893+00:00", "source": "github"}, "name": {"kind": "observed", "observed_at": "2026-08-28T04:06:19.249893+00:00", "source": "github"}, "platform": {"confidence": null, "enriched_at": "2026-08-30T02:50:59.155304+00:00", "kind": "inferred", "prompt_version": 1, "sources": [{"content_hash": "ca7720eea288291430efeea567081fb24397bfa971a6140e1840742d2862cb9d", "fetched_at": "2026-08-28T04:06:19.249893+00:00", "kind": "readme", "missing": false, "url": "https://github.com/Sumandora/remove-refusals-with-transformers"}], "taxonomy_version": 1}, "pushed_at": {"kind": "observed", "observed_at": "2026-08-28T04:06:19.249893+00:00", "source": "github"}, "repo": {"kind": "observed", "observed_at": "2026-08-28T04:06:19.249893+00:00", "source": "github"}, "stars": {"kind": "observed", "observed_at": "2026-08-28T04:06:19.249893+00:00", "source": "github"}, "tags": {"confidence": null, "enriched_at": "2026-08-30T02:50:59.155304+00:00", "kind": "inferred", "prompt_version": 1, "sources": [{"content_hash": "ca7720eea288291430efeea567081fb24397bfa971a6140e1840742d2862cb9d", "fetched_at": "2026-08-28T04:06:19.249893+00:00", "kind": "readme", "missing": false, "url": "https://github.com/Sumandora/remove-refusals-with-transformers"}], "taxonomy_version": 1}, "topics": {"kind": "observed", "observed_at": "2026-08-28T04:06:19.249893+00:00", "source": "github"}, "urls": {"kind": "observed", "observed_at": "2026-08-28T04:06:19.249893+00:00", "source": "github"}, "use_cases": {"confidence": null, "enriched_at": "2026-08-30T02:50:59.155304+00:00", "kind": "inferred", "prompt_version": 1, "sources": [{"content_hash": "ca7720eea288291430efeea567081fb24397bfa971a6140e1840742d2862cb9d", "fetched_at": "2026-08-28T04:06:19.249893+00:00", "kind": "readme", "missing": false, "url": "https://github.com/Sumandora/remove-refusals-with-transformers"}], "taxonomy_version": 1}, "what_it_is": {"confidence": null, "enriched_at": "2026-08-30T02:50:59.155304+00:00", "kind": "inferred", "prompt_version": 1, "sources": [{"content_hash": "ca7720eea288291430efeea567081fb24397bfa971a6140e1840742d2862cb9d", "fetched_at": "2026-08-28T04:06:19.249893+00:00", "kind": "readme", "missing": false, "url": "https://github.com/Sumandora/remove-refusals-with-transformers"}], "taxonomy_version": 1}, "when_to_avoid": {"confidence": null, "enriched_at": "2026-08-30T02:50:59.155304+00:00", "kind": "inferred", "prompt_version": 1, "sources": [{"content_hash": "ca7720eea288291430efeea567081fb24397bfa971a6140e1840742d2862cb9d", "fetched_at": "2026-08-28T04:06:19.249893+00:00", "kind": "readme", "missing": false, "url": "https://github.com/Sumandora/remove-refusals-with-transformers"}], "taxonomy_version": 1}, "when_to_choose": {"confidence": null, "enriched_at": "2026-08-30T02:50:59.155304+00:00", "kind": "inferred", "prompt_version": 1, "sources": [{"content_hash": "ca7720eea288291430efeea567081fb24397bfa971a6140e1840742d2862cb9d", "fetched_at": "2026-08-28T04:06:19.249893+00:00", "kind": "readme", "missing": false, "url": "https://github.com/Sumandora/remove-refusals-with-transformers"}], "taxonomy_version": 1}}, "score": {"components": {"activity": 54, "longevity": 59, "rhythm": 35}, "computed_at": "2026-09-02T17:46:02.011165+00:00", "flags": ["no_releases"], "formula": "round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)", "inputs": {"age_days": 839, "days_push": 279, "days_rel": null, "gap_med": null, "n_releases_24m": 0}, "score": 48, "version": 2}, "staleness": {"enrichment_outdated": false, "low_confidence": false, "scrape_days": 9, "stale_scrape": false}}