Research & Papers

Biohub、DOE、NIHなどが18億ドル規模の投資を行い、AIモデル訓練用のオープンで標準化された生物データを生成する取り組みを開始した

• •Biohub、DOE、NIHおよびパートナーが18億ドルの世界的なコミットメントを行った
• •生成されるデータはオープンかつ標準化されており、AIモデルの訓練に使用される
• •これらのAIモデルは、介入に対する細胞の反応を予測することを目的としている

Many proofs presented by OpenAI deviate from guidelines established by a group of mathematicians who advised the lab.

• •OpenAI's presented proofs deviated from guidelines set by a group of mathematicians who advised the lab.
• •The article points out that OpenAI's mathematical solutions do not meet current field standards.
• •OpenAI is collaborating with a group of mathematicians to advance its research.

SamsungLabs released official implementations of LittleBit and LittleBit-2, introducing LLM compression techniques.

• •It includes the official implementation of LittleBit, presented at NeurIPS 2025.
• •It also includes the implementation of LittleBit-2, the successor version presented at ICML 2026.
• •The code is publicly available on GitHub by SamsungLabs.

News reported that OpenAI retracted three mathematics papers.

Claude Codeを活用することで、これまで存在が知られていなかった惑星を発見した可能性があると主張する記事

OpenAIがGitHubの数学リポジトリを更新し、6つの新しいLean形式化、19件の修正、3件の取り下げを実施した

• •リポジトリには6つの新しいLean形式化と19件の修正が追加された
• •3つの数学的結果が取り下げられた
• •トップラインの結果の約42%が形式化済みとなった

An article critiques OpenAI's Math 1.0 and argues for Math 2.0, which evaluates mathematical progress beyond just problem-solving.

• •Math 1.0 prioritized solving unsolved problems, but its optimization has reached an unsustainable stage.
• •Math 2.0 proposes reducing the focus on live problem-solving in favor of exposition, community building, and exploring new research fields.
• •The author believes AI can contribute positively to all these aspects of mathematical progress.

Explains ten misconceptions about physical AI identified by Gartner, offering a critical view on the advantage of manufacturing data and tacit knowledge.

• •Gartner states it is a misconception that Japanese companies hold an advantage in physical AI due to manufacturing data and tacit knowledge.
• •Organizes specific arguments regarding ten misconceptions about physical AI.
• •Conveys that possessing data or on-site knowledge alone does not guarantee a competitive advantage.

Rumors spread that a robot solved a mathematical problem being worked on by complexity theory researcher Dana Moshkovitz, sparking discussion within her household.

• •Complexity theory researcher Dana Moshkovitz was working on a specific mathematical problem.
• •Rumors spread that a robot had solved the problem.
• •Her nine-year-old son told his researcher mother about the rumor and teased her.

A process for automatically converting AI-generated mathematical text into formal languages for verification may lack reliability in proving Navier-Stokes blow-up.

• •Automatic formalization is used to verify AI-generated mathematical texts, such as the Navier-Stokes blow-up proof announced by OpenAI.
• •In this process, AI systems translate from natural language into formal languages like Lean to enable mechanical verification.
• •The article argues that this translation process does not provide sufficient reliability for mathematical proofs.