brookszdblog

  Is it really a “six-sided warrior”? ByteDance's new model Seed3D 1.0 gets put to the test! It must be said, China's large models are truly showing a diverse and vibrant landscape. On October 23rd, ByteDance's Seed team unveiled the next-generation 3D generative large model Seed3D 1.0 ! Officially described as a 3D foundational model, it combines physical simulation accuracy with scalability, enabling “end-to-end generation from a single 2D image to photorealistic 3D assets.” In other words, without complex operations, ordinary users can quickly obtain a complete 3D model featuring detailed geometry, realistic textures, and physically based rendering (PBR) materials using just a single photo. (Video from ByteDance's official website: https://seed3d.dev/ ) Built on the innovative Diffusion Transformer architecture and trained on massive datasets, this model ensures multi-view consistency and material realism in 3D generation. Additionally, 3D models generated by S...

  RL infrastructure is difficult to standardize like cloud services. oversea: Is RL currently the most crucial element for building agents? In your practical work, where do you see the most significant improvements and areas requiring deeper exploration? Macaron Optimizes Memory with RL Kaijie Chen:   Most companies can't handle RL. We initially ran RL on a 70B model, but its memory capabilities for writing fiction were insufficient—it required reinforcement to produce 100,000 or 200,000 words. Later, our team continued exploring. This year, following the r1 paradigm—especially after Deepseek's 0528 release—we migrated RL from a 70B model to a 671B-scale model. In China, very few teams can independently develop RL on 671B-scale models—probably fewer than five. Even many teams capable of pre-training, like Zhipu, cannot handle RL on dense 671B models. Most companies working on RL remain within the 10-200B range, where 200B is a watershed.   Models under 200B can still be t...