DMax is a new paradigm for efficient diffusion language models (dLLMs) that enables aggressive decoding parallelism while preserving generation quality. This repository contains the DMax-Math-16B model, specialized for highly parallel math and reasoning tasks. — Aggressive Decoding Parallelism: Achieves 6.0 TPF on math and reasoning tasks and 6.6 TPF on code tasks while preserving accuracy. — Self-Revising dLLM: Extends a pretrained MDLM into a UDLM with an intrinsic ability to revise its own erroneous predictions during decoding. — Soft Parallel Decoding: Uses interpolation between mask and token embeddings to propagate confidence priors from previous steps. Superior Parallelism-Accuracy Trade-off, Increased TPF with Maintained Accuracy.
Модальности:
Генерация текста
Области применения:
Диалог / чат Математика
Задача: Генерация текста
Автор: Zigeng
Теги: llada2_moe, conversational, custom_code
Лайков: 4 | Загрузок: 30
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.