{"slug":"openlair-transformer-lens-interpretability","source_name":"openlair/transformer-lens-interpretability","name":"Openlair/Transformer Lens Interpretability","description":"Provides guidance for mechanistic interpretability research using TransformerLens to inspect and manipulate transformer internals via HookPoints and activation caching. Use when reverse-engineering model algorithms, studying attention patterns, or performing activation patching experiments.","version":1,"lift":{"pass_rate_delta_pts":22.73,"pass_rate_pct":100,"total_cases":22,"passed_cases":22,"tokens_delta_pct":153.8,"turns_delta_pct":0,"verdict":"pass","benchmark_model":"gemini-3.6-flash","grading_method":"judged","completed_at":"2026-08-08T01:46:24.167128+00:00"},"skill_score":1,"benchmark_models":[{"model":"gemini-3.6-flash","headline":true,"delta_pts":22.73,"with_pass_pct":100,"without_pass_pct":77.3,"tokens_delta_pct":153.8,"turns_delta_pct":0,"total_cases":22,"cases_aggregated":22,"verdict":"pass","never_hurt":true,"completed_at":"2026-08-08T01:46:24.167128+00:00","run_id":"9e8cacca-9fd1-46e2-a7b4-0a580d222f4c","version_number":1,"is_latest_version":true,"gate":null}],"trust":{"skill_safety":"passed","safety_status":"clean","intent_verdict":"safe","content_status":"clean","indexable":true},"license":"MIT","install_count":0,"manifest_hash":"d16303ff65311bfd90f3e3abec795adfc5054bc4eeb8849f821be0a28f5af649","raw_url":"https://app.decimal.ai/s/openlair-transformer-lens-interpretability/SKILL.md","scorecard_url":"https://app.decimal.ai/skills/openlair-transformer-lens-interpretability"}