← back to paper

Review history

arxiv: 2608.00434 · 2 revisions

AdaMTP: An Adaptive Training Paradigm for Multi-Token Prediction

1. 2026-08-15 CONDITIONAL MODERATE v1.4.0-alltime-deepseek-medium novelty 7.0
   87468 ms 13262 in 11116 out 2026-08-15T15:19:10.886698+00:00
2. 2026-08-05 REJECT MODERATE v1.3.0-alltime-deepseek novelty 6.0
   157232 ms 13251 in 15129 out 2026-08-05T00:59:22.189736+00:00