以DeepSeek‑R1为例，仅靠强化学习训练，模型在AIME数学推理基准上的pass@1从15.6%提升至 77.9%，充分展示了RL在低数据量条件下即可实现大幅能力跃升，迅速成为后训练赛道的新范式。

新智元报道 ...

The bug allows attacker-controlled model servers to inject code, steal session tokens, and, in some cases, escalate to remote ...

Aider is a “pair-programming” tool that can use various providers as the AI back end, including a locally running instance of Ollama (with its variety of LLM choices). Typically, you would connect to ...

Describing AI development as an "arms race" might seem needlessly bombastic, but there's a reason why this term has entered common usage. It encapsulates the speed and intensity at which companies are ...

Z.ai released its complete model weights under an MIT license, allowing developers to download and run it locally—completely eliminating ongoing API costs.

There’s pretty much always a frenzied news cycle of some kind around OpenAI. Fortune’s tech team recently dove into what’s going on behind the scenes, in a feature published this week and helmed by ...

Sam Altman’s decision to declare a “code red” at OpenAI earlier this month may have caught the industry’s attention, but it wasn’t a first for the artificial intelligence company. The San ...

Anthropic said on Wednesday it would release its Agent Skills technology as an open standard, a strategic bet that sharing its approach to making AI assistants more capable will cement the company's ...

OpenAI released a new flagship image generation model today as it moves to counter recent concerns that it is slipping behind rivals in the race to capture both consumer and business mindshare. The ...

OpenAI is rolling out a new version of ChatGPT Images that promises better instruction-following, more precise editing, and up to 4x faster image generation speeds. The new model, dubbed GPT Image 1.5 ...