Keep on to blur preview images; turn off to show them clearly
RT @YouJiacheng: Unfortunately we can't further scale k of pass@k after we scale RLVR compute by 10-1000×. Because AIME only has 1000 answe…
We're in a race. It's not USA vs China but humans and AGIs vs ape power centralization. @deepseek_ai stan #1, 2023–Deep Time «C’est la guerre.» ®1
RT @Yuchenj_UW: Whenever people ask me, “Is Muon optimizer just hype?” I need to show them this. Muon isn’t just verified and used in Kim…
We're in a race. It's not USA vs China but humans and AGIs vs ape power centralization. @deepseek_ai stan #1, 2023–Deep Time «C’est la guerre.» ®1
Ok another easter eggs (I might have enough till christmas): we obviously made sure 56M Monad knows quite a few things about Tiny Recursive Model, the groundbreaking 7M model from @jm_alexia
Reasoning models have come! Co-founder @pleiasfr
This can only happen if the US bans imports of normal robots from a normal robotic superpower (Chyna), and will lead to a further widening of industrial capability gap. These things should be made disposable commodities, not «serious property» akin to slaves in the South.
We're in a race. It's not USA vs China but humans and AGIs vs ape power centralization. @deepseek_ai stan #1, 2023–Deep Time «C’est la guerre.» ®1
RT @eddybuild: today, we’re open sourcing the largest egocentric dataset in history. - 10,000 hours - 2,153 factory workers - 1,080,000,00…
百度刚发了个新VLM模型，给大家整理下： ERNIE-4.5-VL-28B-A3B-Thinking 多模态模型，这个模型虽然总参数有 28B，但采用了 MoE 架构，实际激活参数只有 3B，的确是轻量级。 关键是性能还挺强，在多个基准测试上已经接近行业顶级旗舰模型的表现了。 这个模型的主要特色包括： 视觉推理：这个之前给大家演示过不少，不但能识别图片元素还能推理出内容。 STEM 推理：这个是有用的，可以处理数学、物理等拍照解题任务 视觉定位：提供更精确的定位能力和灵活的指令执行 带图思考：模型可以像人一样思考，能够自由放大缩小图片来把握每一个细节 工具调用：可以使用图片搜索等功能来识别长尾知识 -视频理解：识别视频不同时间段的内容变化 目前部署也支持很多，包括 vLLM 等。不过官方并没有放出 space demo 供大家直接测试，可以等一等。如果大家感兴趣的话稍后我会做个测试。 模型地址：
详细数据
