diff options
Diffstat (limited to 'BABYAI_SHARED_FEEDBACK.md')
| -rw-r--r-- | BABYAI_SHARED_FEEDBACK.md | 10 |
1 files changed, 10 insertions, 0 deletions
diff --git a/BABYAI_SHARED_FEEDBACK.md b/BABYAI_SHARED_FEEDBACK.md index 9443253..009709f 100644 --- a/BABYAI_SHARED_FEEDBACK.md +++ b/BABYAI_SHARED_FEEDBACK.md @@ -184,3 +184,13 @@ history lengths four and eight for 25 epochs, using only BP and clean KP. The same `70%` BP, `60%` clean-KP, and 20-point BP mission-lesion thresholds apply; ties favor the shorter history. This changes the missing state information, not the shared-feedback signal or its strength. + +## P1 outcome + +Four-step history passed: BP reached `77.6%` rollout success, clean KP reached +`61.6%`, and removing the mission reduced BP success to `44.6%`. Eight-step +history did not pass (`67.2%` BP and `42.2%` clean KP), so the frozen selector +chooses four steps. The P2 mechanism endpoint uses four hidden layers, width +256, learning rate `0.03`, four-step history, 40 epochs, and model/minibatch +seeds `4101`, `4102`, and `4103`. No raw or SDIL result was run or read during +selection. |