1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
|
{
"checks": {
"all_control_invariants_audited": true,
"dynamic_reference_is_finite": true,
"matched_degrades_or_collapses_by_at_least_3_points": true,
"raw_and_matched_use_identical_source": true,
"raw_degrades_or_collapses_by_at_least_5_points": true,
"two_frozen_controls_present": true
},
"claim": "Under four-RMS soma-predictable apical traffic, innovation subtraction is necessary relative to both raw and norm-matched KP controls.",
"metrics": {
"analysis_source_commit": "ff0069f8d0dabbb8819a726bd899381e8a95462c",
"clean_kp_accuracy": 0.9126,
"control_mean_wall_s": 8237.62871158123,
"dynamic_accuracy": 0.9118,
"dynamic_minus_matched_points": 81.18,
"dynamic_minus_raw_points": 81.18,
"matched_accuracy": 0.1,
"matched_finite": false,
"prerequisite_sha256": "48957573e6feb18e5c2dd0c6785bbd20f37126dd33e8bb7cb357190d236b4937",
"raw_accuracy": 0.1,
"raw_finite": false,
"training_source_commit": "31b70e1b1f2e7fd26c9c3228f525fc0b5192cd81"
},
"protocol": "kp_teaching_signal_ablation_validation_v1",
"scope_limit": "This is a conditional robustness result on one validation seed; it is not clean-setting or strong-baseline superiority.",
"status": "passed"
}
|
