{"slug":"perminder-klair-subwave-llm-bench","source_name":"perminder-klair/subwave-llm-bench","name":"Perminder Klair/Subwave LLM Bench","description":"Benchmark and compare LLM models for SUB/WAVE's on-air calls — track picks, segments, listener requests, DJ scripts, banter, and programme beats — in both candidate-pool and agent modes, using controller/scripts/llm-bench (the matrix harness) or the legacy picker-only controller/scripts/picker-test.mjs. Use this skill whenever the user wants to assess, benchmark, compare, or test which LLM model to run the station on — phrases like \"which model should I use\", \"benchmark the picker / the DJ / thi","version":2,"lift":{"pass_rate_delta_pts":58.33,"pass_rate_pct":87.5,"total_cases":24,"passed_cases":21,"tokens_delta_pct":95.8,"turns_delta_pct":0,"verdict":"mixed","benchmark_model":"gemini-3.6-flash","grading_method":"judged","completed_at":"2026-09-29T05:56:21.737916+00:00"},"skill_score":0.875,"benchmark_models":[{"model":"gemini-3.6-flash","headline":true,"delta_pts":58.33,"with_pass_pct":87.5,"without_pass_pct":29.2,"tokens_delta_pct":95.8,"turns_delta_pct":0,"total_cases":24,"cases_aggregated":21,"verdict":"mixed","never_hurt":false,"completed_at":"2026-09-29T05:56:21.737916+00:00","run_id":"98060925-ed11-42fa-ab9c-ad12159ab2df","latest_attempt":{"id":"98060925-ed11-42fa-ab9c-ad12159ab2df","runner_model":"gemini-3.6-flash","overall_verdict":"mixed","completed_at":"2026-09-29T05:56:21.737916+00:00","gate":null},"version_number":1,"is_latest_version":false,"gate":null},{"model":"gemini-3.8-flash","headline":false,"delta_pts":null,"with_pass_pct":null,"without_pass_pct":null,"tokens_delta_pct":29.6,"turns_delta_pct":0,"total_cases":23,"cases_aggregated":22,"verdict":"mixed","never_hurt":true,"completed_at":"2026-10-21T20:24:55.071748+00:00","run_id":"474f0bd9-e1e4-4f85-b0a3-6198851ba135","latest_attempt":{"id":"474f0bd9-e1e4-4f85-b0a3-6198851ba135","runner_model":"gemini-3.8-flash","overall_verdict":"mixed","completed_at":"2026-10-21T20:24:55.071748+00:00","gate":"not_comparable"},"version_number":2,"is_latest_version":true,"gate":"not_comparable"}],"trust":{"skill_safety":"passed","safety_status":"clean","intent_verdict":"safe","content_status":"clean","indexable":true},"license":"MIT","install_count":0,"total_uses":0,"total_users":0,"manifest_hash":"f91b99c4a34d296db3d828d54e8efdb96438de1f0a16c471712178bb0cc40d2a","install_name":"perminder-klair-subwave-llm-bench","version_id":"0281300c-b7b2-448d-8b42-8101dc0783f0","body_sha256":"f91b99c4a34d296db3d828d54e8efdb96438de1f0a16c471712178bb0cc40d2a","render_hash":"d13449362f016f0aa328bd971cfe44101495baf94ef3d71753045733667a99ba","raw_url":"https://app.decimal.ai/s/perminder-klair-subwave-llm-bench/SKILL.md","scorecard_url":"https://app.decimal.ai/skills/perminder-klair-subwave-llm-bench"}