whiskeywhiskey's picture
Publish WALDO experimental reference artifacts
c1ac40f verified
Raw History Blame Contribute Delete
1.08 kB
{
"recorded_at": "2026-10-31T21:25:00.141786+00:00",
"status": "completed-and-reviewed",
"training_complete": true,
"native_run_id": "5652ba3bce3a7f6c",
"selected_step": 271073,
"final_step": 295716,
"final_quality_qualification_passed": false,
"gates": {
"finite_metrics": true,
"artifact_reload_exact": true,
"native_precision_tolerance": true,
"selected_in_final_twenty_percent": true,
"bpb_improves_over_midpoint": true,
"bpb_at_most_080": true,
"two_sentences": true,
"sampled_loops": true,
"greedy_loops": true,
"manual_grammar_topic_18_of_20": false
},
"manual_clear_passes": 11,
"manual_borderline": 4,
"manual_generous_upper_bound": 15,
"thresholds_unchanged": true,
"conclusion": "Successful full-plan systems run; selected candidate fails the frozen final behavioral qualification. Do not label it a promoted rung or a qualified assistant.",
"reviewer": "Assistant manual review; not independent or blinded.",
"training_stop": "Native planned completion, not a discretionary review pause."
}