As coding models improve, evals need to become harder, fairer, and more trustworthy. Better benchma... - 惯性聚合

推荐订阅源

Full Disclosure

Stack Overflow Blog

Help Net Security

Microsoft Azure Blog

The GitHub Blog

cs.CV updates on arXiv.org

cs.AI updates on arXiv.org

Recorded Future

Y Combinator Blog

Threat Intelligence Blog | Flashpoint

Schneier on Security

KPMG report finds enterprise disconnect between AI and its ROI | CIO

Cybersecurity and Infrastructure Security Agency CISA

TaoSecurity Blog

Security Latest

News and Events Feed by Topic

Hacker News - Newest: "LLM"

DataBreaches.Net

Security Archives - TechRepublic

Hacker News: Front Page

aimingoo的专栏

Recent Commits to openclaw:main

Vulnerabilities – Threatpost

LINUX DO - 最新话题

cs.CL updates on arXiv.org

Google Online Security Blog

Darknet – Hacking Tools, Hacker News & Cyber Security

About on SuperTechFans

www.infosecurity-magazine.com

CXSECURITY Database RSS Feed - CXSecurity.com

Schneier on Security

News and Events Feed by Topic

有赞技术团队

大猫的无限游戏

The Blog of Author Tim Ferriss

Security Affairs

Know Your Adversary

Simon Willison's Weblog

Google DeepMind News

The Cloudflare Blog

OpenAI(@OpenAI)

OpenAI (@OpenAI) on X OpenAI (@OpenAI) on X OpenAI (@OpenAI) on X ChatGPT Work is powered by GPT-5.6. GPT-5.6 makes ChatGPT state of the art at reasoning through com... On web and mobile, ChatGPT Work is rolling out today for Pro, Enterprise, and Edu plans. It will rol... ChatGPT Work reflects a shift in how people are using AI, moving beyond just answering questions to ... Introducing ChatGPT Work, a new agent in ChatGPT powered by Codex and GPT-5.6. It can take action a... GPT‑5.6 is available starting today across ChatGPT, Codex, and the OpenAI API. The rollout is starti... GPT‑5.6 improves artifact quality across presentations, documents, and spreadsheets, and works bette... OpenAI (@OpenAI) on X OpenAI (@OpenAI) on X GPT-5.6 Sol sets a new standard for intelligence and efficiency, delivering state-of-the-art perform... OpenAI (@OpenAI) on X OpenAI (@OpenAI) on X OpenAI (@OpenAI) on X GPT-Live is now fully rolled out to all ChatGPT users on Go, Plus, and Pro plans. Free user rollout ... To audit SWE-Bench Pro, we used model-based investigator agents alongside independent reviews from f... We audited SWE-Bench Pro, one of the most widely used AI coding benchmarks, and found it no longer r... Our audit of SWE-Bench Pro found that a meaningful share of public tasks contain issues that can dis... To audit SWE-Bench Pro, we used model-based investigator agents alongside independent reviews from f... As coding models improve, evals need to become harder, fairer, and more trustworthy. Better benchma... We audited SWE-Bench Pro, one of the most widely used AI coding benchmarks, and found it no longer r... For questions that require web search, deeper reasoning, or more complex work, GPT-Live can delegate... These new models are rolling out to everyone over the next few days, starting today. They’ll be av... GPT-Live is built on a full-duplex architecture, meaning it can listen and speak at the same time. Introducing GPT-Live, a new generation of voice models for natural human-AI interaction. Rolling ou... GPT-Live makes talking with AI feel like having a real conversation. It’s also our smartest voice mo... The next generation of ChatGPT Voice is here. Livestream starts at 10am PT. https://t.co/giS4K1yNh9 Listen up. Livestream at 10am. GPT-5.6 Sol, along with Terra and Luna, will launch publicly this Thursday. We’re expanding preview... We’re introducing GeneBench-Pro, a research-level benchmark for a harder kind of AI progress: how we... GPT-5.6 Sol is our most capable model yet for cybersecurity. It shifts the performance-efficiency f... GPT‑5.6 Sol launches with our most robust safety stack yet. We strengthened real-time protections a... GPT‑5.6 Sol sets a new state of the art on Terminal‑Bench 2.1, which tests complex command-line work... Sol is our new flagship and a step function better than GPT-5.5. Terra delivers performance competi... Introducing a limited preview of GPT-5.6 Sol, our next generation frontier model, as well as GPT-5.6... We believe in broad access and plan to make GPT-5.6 Sol, Terra, and Luna generally available in the ... https://t.co/C0hhXzOmSP Work at OpenAI is being transformed by agents, in every department. Across our entire company, peop... We have a new version of GPT-5.5 Instant for you, and it's much more fun to talk to. Our most-used ... https://t.co/FlbllRaSeb We’ve designed and built our first AI chip: Jalapeño. Designed from the ground up by OpenAI and bro... OpenAI DevDay 2026 applications are now open! Our biggest developer event gets even bigger. 📍 San... You can also tune in from home. The opening keynote will be livestreamed on September 29. Patch the Planet is our effort to help open source maintainers move from security findings to merged... We’re also launching the OpenAI Daybreak Cyber Partner Program with leading security software and se... We’re expanding OpenAI Daybreak to help democratize patching vulnerable software at machine speed: ... GPT-5.5-Cyber is our most capable cyber model yet, designed for advanced, authorized defensive work:... Codex Security empowers defenders with out-of-the-box security workflows. Teams can run deep scans,... We also tested whether alignment persisted under pressure. The model was harder to steer toward ha... This is an early step toward more robustly beneficial and aligned models: training models to carry b... A small amount of this data produced broad gains beyond the training scenarios. Compared with a com... We trained models with reinforcement learning on realistic conversations to reinforce beneficial tra... As AI takes on longer, higher-stakes tasks, we want models to carry beneficial and safe behavior int... Improving human health will be one of the most personal, tangible impacts of AGI. As our models con... GPT-5.5 Instant is now on par with our frontier Thinking models for health-related questions. Every... To improve our models, we collaborate with a global network of hundreds of physicians across 60 coun... Rare disease diagnosis is challenging, as sequencing can surface millions of variants, and medical k... Many of these cases had evaded years of expert analysis. This study suggests AI could make expert-l... Together with researchers at Boston Children’s Hospital and Harvard, we published a study in NEJM AI... Introducing LifeSciBench, a benchmark for measuring and improving how well AI supports real-world li... Benchmarks often test biological knowledge or narrow skills. The tasks in LifeSciBench test whether ... LifeSciBench is a foundation for more realistic evaluation, targeted improvements, and continued par... The full process took about 2.5 months, plus another half month for human chemists to write up the r...

As coding models improve, evals need to become harder, fairer, and more trustworthy. Better benchma...

OpenAI (@Ope · 2026-07-09 · via OpenAI(@OpenAI)

此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。原文来自 — 版权归原作者所有。