Results / atlas/qwen3-0.6b-4bit/interventions/v1/mapcard.json
{
"map_id": "atlas/qwen3-0.6b-4bit/interventions/v1",
"map_type": "interventions",
"model_id": "mlx-community/Qwen3-0.6B-4bit",
"model_hash": "392e8d466d56100ada00eb82031fb854297fc9e389b7d303eba3af114e87bce2",
"quantization": "q4 (mlx)",
"commit": "f7c298d2859ed43654d388618b7274266e7e562b",
"config": "results/expC_causal_verification/20260812T071954Z/results.json",
"created": "2026-08-12",
"hardware_manifest": {
"author": "Simon-Pierre Boucher",
"contact": "contact@spboucher.ai",
"website": "https://modelmap.io",
"chip": {
"brand": "Apple M5 Max",
"cores_total": 18,
"cores_performance": 6,
"cores_efficiency": 12
},
"memory": {
"unified_gb": 48,
"pagesize": 16384
},
"os": {
"system": "Darwin",
"version": "27.0",
"arch": "arm64"
},
"software": {
"python": "3.14.4",
"numpy": "2.5.2",
"mlx": "0.32.0",
"torch": "2.13.0",
"safetensors": "0.8.0"
}
},
"confidence_level": 2,
"regenerate_command": ".venv/bin/python experiments/micro/expC_causal_verification/implementation/benchmark_v4.py && .venv/bin/python experiments/micro/expC_causal_verification/implementation/make_interventions_mapcard.py",
"seeds": [
780,
"Ahalf1",
"Ahalf2",
"bootA",
"Bhalf1",
"Bhalf2",
"bootB"
],
"prompt_sets": [
"agreement_A halves (direction est.)",
"agreement_B halves (direction est.)",
"fresh held-out minimal-pair bank (8 unseen locations)"
],
"controls": [
"random-direction erasure per layer (3 dirs, netted out)",
"six-source replication with pre-registered bar (runs #3-#4)",
"late band excluded as estimator-unstable (run #3 refusal)"
],
"methods_in_agreement": [
"difference-in-means probing (direction exists, decodable)",
"direction erasure (causally load-bearing)"
],
"interventions": [
"rank-1 direction erasure at each layer's output, all positions"
],
"replication_rate": 0.7254,
"per_dataset_agreement": null,
"ablation_schemes": [],
"featurizer_class": "linear (difference-in-means direction)",
"intervention_protocol": "erase h' = h − ⟨h−μ,u⟩u at layer ℓ; logit-margin metric; random-direction null netted out; BAND granularity",
"negative_result": false,
"notes": "Level 2, NOT 3: the two agreeing methods share the diff-of-means estimator; activation-addition steering is the registered Level-3 path. This entry exists because run #3's per-layer version was REFUSED by the gate — the band is the granularity that replicates. Anti-correlates with probes/v2 layer ranking (survival ledger 0/2): decodability peaks ≠ causal joints.",
"author": "Simon-Pierre Boucher",
"contact": "contact@spboucher.ai",
"website": "https://modelmap.io",
"schema_version": "0.1"
}