Shanghai AI Laboratory

Headlines · 2

1. 2d
   Intern-S2 helps prove 3D Ghys conjecture, checked in Lean

   Shanghai AI Laboratory said on October 15 that Xu Disheng, a visiting scholar there and professor at Greater Bay Area University, used its Intern-S2 science multimodal model in a human-machine collaboration to prove the three-dimensional Ghys conjecture, a problem open for more than 30 years. A Peking University team led by Dong Bin and Liu Ruochuan formalized all four main theorems with its own SHEAF framework and checked them in Lean, taking about five days with up to 100 agents in parallel and roughly 550,000 lines of Lean code. The lab says Intern-S2 ran seven successive rounds of reasoning, expanding its problem graph from 15 to 389 nodes and producing a 400-page first-round derivation report before the final thousand-plus pages; the paper is on arXiv and the formalization code on GitHub. The lab also claims the model leads existing open models on IMO-Proof and AdvancedMathBench and beats Gemini 3.1 Pro on some metrics, and notes an expert had tried the problem with GPT-6 without finding a proof.

   IT之家 AI · 🔥 10
2. 3d
   Intern Lumina U2: multi-codebook diffusion LLM for unified vision and image generation

   The paper introduces Intern Lumina U2, which the authors describe as the first diffusion language model to represent multi-codebook visual tokens in a fully discrete space, where each visual position is encoded by a group of codes drawn from complementary codebooks to raise representational capacity while staying compatible with discrete text modeling. At the input the embeddings are concatenated per position, while at the output a lightweight multi-codebook autoregressive head predicts codes within each position sequentially and the diffusion backbone processes all positions in parallel. Built on AToken's unified 4D representation, it handles image, video and 3D inputs in a single backbone and, according to the authors, achieves leading results among unified models in visual understanding, image generation and image editing; a stability-aware sampling strategy combining confidence-plateau-guided unmasking with drift-triggered cache refreshing is reported to speed up image generation by 4.45x. The authors also say they built a full-stack training, inference and evaluation system on both Ascend and NVIDIA platforms to verify cross-platform effectiveness.

   Hugging Face · Papers · 🔥 0

Experience and discussion from the community

Share my Shanghai AI Laboratory experienceAsk about Shanghai AI Laboratory