开启时会模糊预览图，关闭后正常显示
Felt like claude artifacts to me but maybe a little better it went back to claudey websites with lots of neon and purples which was a bit disheartening..
Cofounder and Head of Post Training @NousResearch, prev @StabilityAI Github: https://t.co/LZwHTUFwPq HuggingFace: https://t.co/sN2FFU8PVE
RT @_thomasip: in a sea of people chasing max profit and clout, there are those who genuinely want to move the field forward.
Cofounder and Head of Post Training @NousResearch, prev @StabilityAI Github: https://t.co/LZwHTUFwPq HuggingFace: https://t.co/sN2FFU8PVE
Also, at least use their sora 2 model instead of veo or w/e slop that is if you want to make ai videos of sama
Cofounder and Head of Post Training @NousResearch, prev @StabilityAI Github: https://t.co/LZwHTUFwPq HuggingFace: https://t.co/sN2FFU8PVE
mmm they may not be #1 but I don't think they are 95th at anything lol please stick to facts so I dont have to defend openai
Also, at least use their sora 2 model instead of veo or w/e slop that is if you want to make ai videos of sama
I think this is generally fine except the part about having to constantly remind them the ai is not a human - seems like that'd be really annoyingly immersion breaking. It's not clear to me what an ai companion is either though is chatgpt one or only characterai or even more narrow only replika?
Cofounder and Head of Post Training @NousResearch, prev @StabilityAI Github: https://t.co/LZwHTUFwPq HuggingFace: https://t.co/sN2FFU8PVE
I dont know why OS Inference Engines are still having this problem across like.. every model at varying rates - this is GLM 4.6 and there's 23/10000 samples here that are like this - through openrouter, so either vllm or sglang... The problem's been happening for 2 years now, meanwhile gpt/claude etc I have only seen this kind of thing with maybe 2 or 3 times across 500k+ responses..
Cofounder and Head of Post Training @NousResearch, prev @StabilityAI Github: https://t.co/LZwHTUFwPq HuggingFace: https://t.co/sN2FFU8PVE