https://cognaptus.com/tags/multi-token-prediction/ Recent content in Multi-Token-Prediction on Cognaptus Hugo -- 0.145.0 en-us Sat, 26 Sep 2026 00:00:00 +0000 https://cognaptus.com/blog/2026-09-26-grow-while-you-roll-training-the-draft-head-inside-the-rl-run/ Sat, 26 Sep 2026 00:00:00 +0000 https://cognaptus.com/blog/2026-09-26-grow-while-you-roll-training-the-draft-head-inside-the-rl-run/ GrowMTP shows that speculative decoding can be learned and amortized inside an LLM reinforcement-learning run, but its experiments also show why acceptance rate is the wrong metric to govern the accelerator alone.