Non-record: CoDA-GQA Differential Attention — First Differential Attention Submission (val_bpb=1.1580)#932

Open

anthony-maio wants to merge 4 commits intoopenai:mainfrom

anthony-maio:submission/coda-gqa-parameter-golf

Commits on Mar 27, 2026

Non-record: CoDA-GQA Differential Attention for Parameter Golf

anthony-maio
and
claude
committed
Fix CoDA pairwise rotation broadcast shape

anthony-maio
and
claude
committed
Add 8xH100 ablation results: CoDA vs baseline

anthony-maio
and
claude
committed

Commits on Apr 1, 2026

Fix submission.json compliance: add date, val_loss, seed_results, model bytes
anthony-maio
committed