ARCHIVE近 30 天历史柱高表示当天去重快讯数量
09/10—10/09 有历史数据
- Sharing AI progress in mathematicsLoading… Share We’re releasing a broad range of new mathematical results produced by an internal frontier model . As we look to improve how we share results with the math community, we’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study (opens in a new window) to develop best practices, and we have drawn on their advice and public recommendations (opens in a new window) to inform how we release these results. F
- Our framework for reporting model misalignmentLoading… Share What misalignment examples we’ll report What misalignment examples we’ll report The misalignment examples we’re sharing today How our disclosure process works What each report will include What misalignment examples we’ll report The misalignment examples we’re sharing today How our disclosure process works What each report will include We are sharing a new framework for tracking, investigating, and disclosing instances of model misalignment at OpenAI, along with six reports on une
- On the Navier–Stokes Millennium Prize ProblemLoading… Share The problem The problem The result How we found the proof Concurrent work Progress and responsibility The problem The result How we found the proof Concurrent work Progress and responsibility We’re sharing a solution to the Navier–Stokes existence and smoothness problem, one of the Millennium Prize Problems. This proof, produced by an internal OpenAI system, shows that the dynamics of the Navier-Stokes equations for fluid motion can develop a singularity in finite time. We’re shar
- Research acceleration: The view inside OpenAILoading… Share 1. Coding agents are reshaping daily work for OpenAI researchers 1. Coding agents are reshaping daily work for OpenAI researchers 2. Researchers are writing more code and running more experiments 3. The work researchers use agents for is changing 4. Pacing model development 5. The path ahead Appendix: Our methods for this post 1. Coding agents are reshaping daily work for OpenAI researchers 2. Researchers are writing more code and running more experiments 3. The work researchers u
- GPT-6 Astra: A new generation of intelligenceGPT Astra GPT-6 Astra: A new generation of intelligence A new generation of intelligence Update on September 29, 2026: Learn about OpenAI's latest model: GPT‑6 .1 Sol . Update on September 22, 2026: We are expanding our GPT‑6 family with GPT‑6 Sol and GPT‑6 Luna. Learn more. We’re introducing GPT‑6 Astra, the world’s most intelligent and aligned model. GPT‑6 Astra brings together years of research and big bets across pre-training, reinforcement learning, and alignment. Astra is state-of-the-ar
- How enabling two settings tripled our scores on the ARC-AGI-3 benchmarkLoading… Share ARC-AGI-3 ARC-AGI-3 Agents do best when they remember what they’ve done Conclusion and recommendations ARC-AGI-3 Agents do best when they remember what they’ve done Conclusion and recommendations A sped-up video of GPT‑5.6 Sol attempting to solve puzzles in the ARC-AGI-3 benchmark, with the official harness (left) and our Responses API harness (right), which retains reasoning and enables compaction. On the leaderboard for this game (opens in a new window) , no frontier model sol
- Separating signal from noise in coding evaluationsLoading… Share Methodology Methodology Human-supervised agent review Human annotation campaign Discussion Methodology Human-supervised agent review Human annotation campaign Discussion Accurately measuring our models’ capabilities is important for sound deployment and safety decisions, including decisions under OpenAI’s Preparedness Framework (opens in a new window) . With each model release, we report results for a variety of external and internal benchmarks to track model progress. When eval
- Introducing GeneBench-ProLoading… Share Dataset construction Dataset construction Evaluation and grading Results Dataset construction Evaluation and grading Results Scientific data rarely arrive with instructions. Researchers must decide whether a pattern reflects biology or noise, whether the data can support the question being asked, and how each result should change what they do next. AI agents are increasingly capable of executing complex analyses, but real scientific research also depends not simply on recalling fa
- A near-autonomous AI chemist improves a challenging reaction in medicinal chemistryShare Why the chemistry problem matters Why the chemistry problem matters Connecting GPT-5.4 to Maria AI and Lab What we found Limitations Preparedness What’s next Why the chemistry problem matters Connecting GPT-5.4 to Maria AI and Lab What we found Limitations Preparedness What’s next OpenAI’s work in science is motivated by a simple belief: advanced AI can become a powerful partner for scientists, helping them explore more ideas, connect distant concepts, design better experiments, and accele
- Introducing LifeSciBenchLoading… Share Agentic AI systems are becoming increasingly capable of performing scientific tasks. However, their usefulness to life science researchers depends on how well they handle the complexity of real research. That work rarely looks like a single fact-recall question or a clean prediction problem. Researchers interpret incomplete evidence, reconcile conflicting results, design difficult experiments, troubleshoot assays, evaluate translational risk, and decide what to do next under uncer
- Predicting model behavior before release by simulating deploymentShare Introduction Introduction How Deployment Simulation works How we tested Deployment Simulation Deployment Simulation significantly expands pre-deployment risk assessment Reducing evaluation awareness Tool simulation for agentic trajectories WildChat and external auditing Limitations Conclusion Introduction How Deployment Simulation works How we tested Deployment Simulation Deployment Simulation significantly expands pre-deployment risk assessment Reducing evaluation awareness Tool simulatio
- Dreaming: Better memory for a more helpful ChatGPTLoading… Share How memory has evolved How memory has evolved How we evaluate memory Carrying forward context Following preferences Staying current over time A more scalable foundation for the future How memory has evolved How we evaluate memory Carrying forward context Following preferences Staying current over time A more scalable foundation for the future Today we’re beginning to roll out a more capable and scalable system for synthesizing memory, developed to tackle the staleness, correctness
- An OpenAI model has disproved a central conjecture in discrete geometryLoading… Share The unit distance problem The unit distance problem New techniques from algebraic number theory What this means for mathematics Why this matters The unit distance problem New techniques from algebraic number theory What this means for mathematics Why this matters For nearly 80 years, mathematicians have studied a deceptively simple question: if you place n n n points in the plane, how many pairs of points can be exactly distance 1 1 1 apart? This is the planar unit distance proble
- What Parameter Golf taught us about AI-assisted researchLoading… Share Technical impressions Technical impressions Record track Nonrecord track Takeaways What’s next? Technical impressions Record track Nonrecord track Takeaways What’s next? We launched Parameter Golf to engage and support the machine learning research community in exploring a new, tightly constrained machine learning problem. We wanted the challenge to be interesting enough to reward real technical creativity, while remaining conceptually simple and easy to verify. Participants had t
- Introducing OpenAI Privacy FilterLoading… Share A small model with frontier personal data detection capability A small model with frontier personal data detection capability Model overview How we built it How Privacy Filter performs Limitations Availability Looking ahead A small model with frontier personal data detection capability Model overview How we built it How Privacy Filter performs Limitations Availability Looking ahead Today we’re releasing OpenAI Privacy Filter, an open-weight model for detecting and redacting person