Introspective Diffusion Language Models / 内省扩散语言模型

📰 2026-04-14 17:30 更新

🔸 Introspective Diffusion Language Models / 内省扩散语言模型

🔗 Introspective Diffusion Language Models
🔥 27 points

原文:
Introspective DiffusionLanguage Models 69.6AIME-24 (I-DLM-8B)vs. LLaDA-2.1-mini 43.3 45.7LCB-v6 (I-DLM-8B)vs. LLaDA-2.1-mini 30.4 2.9-4.1xThroughput overLLaDA-2.1-mini at C=64 LosslessBit-for-bit identicalto base AR model

译文:
内省DiffusionLanguage模型69.6AIME-24 (I-DLM-8B) vs. LLaDA-2.1-mini 43.3 45.7LCB-v6 (I-DLM-8B) vs. LLaDA-2.1-mini 30.4 2.9-4.1x在C = 64时吞吐量LLaDA-2.1-mini无损位与基本AR模型相同


自动更新 · 正文抓取 · 双语翻译

Leave a Comment