mixed-messenger-4b / evaluation /pause-for-performance.json
noahmclaughlin's picture
Release Mixed-Messenger specialist: 24.5% lower held-out prediction error
bfee871 verified
Raw History Blame Contribute Delete
469 Bytes
{
"paused_at": "2026-11-15T06:19:57.455731+00:00",
"reason": "User requires meaningful improvement; initial best development gain was only 0.05445 points. Publication paused and development reopened.",
"archived_baseline_run": "runs/paused-initial-qwen3-4b-base-final-de152c4ceb49002d",
"saved_baseline_episode_files": {
"none": 45,
"agent": 25
},
"adapted_model_evaluation_started": false,
"final_scores_used_for_new_training_decisions": false
}