Shanghai AI Lab Intern-S2 solves 3D Ghys conjecture with human-AI collaboration
Xu Disheng's team used Shanghai AI Lab's open science model Intern-S2 in a human-in-the-loop loop to prove the 3D Ghys conjecture, open for 30 years; Peking University formally verified the proof with its SHEAF framework.
Why it matters
First world-class math breakthrough credited to a Chinese model: recursive multi-batch derivation (389-node problem graph, 1000+ pages of reasoning) plus formal machine verification as the acceptance gate.
What you could build with it
A small research team could adopt the demonstrated workflow — the model expands a problem graph from 15 to 389 derivation nodes across batches while humans steer and prune — to attack other long-open problems, producing machine-checkable proof drafts with formal verification as the final gate.
Does it hold up?
Too early to judge: this is a research milestone, not a shipped product; evidence is one team's human-in-the-loop run, and the underlying open Intern-S2 model itself was released Sept 13. No verified GitHub repo for the formal-verification code was found, so none is listed.
Built with Shanghai AI Lab Intern-S2 solves 3D Ghys…
- Chinese researchers solve 3D Ghys conjecture with homegrown AI modelCGTN · English-language coverage: 7-batch recursive derivation, 400-page first-round report, and PKU's formal verification of all core theorems.
- 上海人工智能实验室书生科学多模态大模型助力中国数学家攻克世界级数学难题澎湃新闻 · Report on the breakthrough, the CMO gold-medal backstory of 书生-S1, and the planned encapsulation into the 书生·端砚 scientific-discovery platform.
Learn more
First spotted on article: source.
More AI models
OpenAI publishes 722 AI-generated math manuscripts from an unreleased modelOpenAI released 722 mathematical manuscripts in 372 result families to GitHub under Apache-2.0, produced by an…models · JEV 0.74StepFun Step-5-Preview — open BF16 weights releasedStepFun published the full BF16 weights for its flagship Step-5-Preview agentic model (600B-parameter sparse MoE, 27B…models · JEV 0.73ElevenLabs Eleven v4 + v4 Turbo: new emotive TTS architecture with a 100ms real-time variantElevenLabs released Eleven v4 and Eleven v4 Turbo on September 28: a new text-to-speech architecture with inline…models · JEV 0.72Mistral Large 4 (Le Chonk): 1.05T-param open-weight MoEMistral's new flagship: a 1.05T-parameter mixture-of-experts multimodal model with 49B active params and 1M-token…models · JEV 0.72VoiceStudio: open-source, fully-local ElevenLabs alternative with voice cloning, dubbing and dictation in 646 languagesVoiceStudio is an open-source, fully-local voice AI studio: voice cloning from a 3-second sample, voice design, video…models · JEV 0.71JetBrains Mellum 2.1: 12B MoE coding model trained for agentsJetBrains released Mellum 2.1, a 12B MoE model (2.5B active parameters, 128K context) under Apache 2.0, trained mainly…models · JEV 0.69
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.