AITOPAI热报
Models

让 JEV 输出 ASCII 字符即可变回自回归 LLM

Jerry Liu@jerryjliu0
Post
View the post
Why we picked it

Jerry Liu 发现了一个有趣的方法，让 JEV 模型输出 ASCII 字符就能变回自回归 LLM，挺有意思的。

Isaac Tai 提出一种方法，通过让 JEV 模型输出 ASCII 字符，可以将其重新配置为自回归语言模型。这种方法利用了 JEV 的内部结构，使其能够生成文本。与传统的自回归模型相比，这种转换方法更简单，且不需要额外的训练步骤。

Original · Jerry Liu

you can just turn jev back into an autoregressive LLM by letting it output ascii lol Isaac Tai 🧇 @hi_im_isaac_ "jev can't generate text" But jev has a lot to say! I gave him a couple hundred common english words to choose from + punctuation Your browser does not support the video tag. 🔗 View on Twitter 🔗 View Quoted Tweet 💬 3 🔄 0 ❤️ 7 👀 754 📊 3 ⚡

#JEV#自回归 LLM#Isaac Tai

Keep reading

Recent stories with the same tag or section
ModelsBuildersOfficialSelected14:08
JEV 1.14.0 adds multimodal support, box output and 262K context

The update closes specific gaps analysts flagged weeks earlier, which is a rare case of a vendor visibly acting on public feedback. Developers building image and video pipelines can now get object coordinates directly from the model instead of adding a separate detection step. The 262K context also makes long multilingual document and media workflows viable on a single model.

#JEV#multimodal#context window#multilingual
官方一手歸藏(guizang.ai)@op7418Source
ProductsBuildersOfficialSelected13:11
JEV 1.14.0 adds multimodal support and 262K context

The "box" output type is the quiet highlight: returning object coordinates means JEV can now drive GUI-style automation and visual grounding workflows that text-only setups cannot. Developers building image-editing or agent pipelines can pair the 262K context with the new type to chain several visual steps in one request. Compared with tools that expose multimodality from day one, JEV took two releases to get there, so regression risk in 1.14.0 deserves a quick test.

#JEV#multimodal#coding assistants
官方一手歸藏(guizang.ai)@op7418Source