wassnameandClaudypoo d9fecd503d explain MFV top1 0.77 vs 0.83: it's the word->digit readout debiasing ...

Exhausted the legitimate levers on the current (digit) readout: top1 0.72 (think
64), 0.77 (256, 512 collapses), 0.72 (BMA n_samples=8). The 82.6% required the old
word-first-token readout, replaced deliberately to drop the uneven-first-piece
word prior. 0.773 is the honest ceiling; 0.83 would need reverting the debiasing
(research poison). README note + journal updated.

Co-Authored-By: Claudypoo <288921227+claudypoo@users.noreply.github.com>

2026-06-25 03:59:30 +08:00

..

img

switch showcase to Qwen3-4B: coherent all poles, clean bidirectional steer

2026-06-25 03:02:41 +08:00

2025_clifford_paper.md

clifford paper

2026-05-03 12:32:56 +08:00

RESEARCH_JOURNAL.md

explain MFV top1 0.77 vs 0.83: it's the word->digit readout debiasing

2026-06-25 03:59:30 +08:00