Get the latest tech news

GPT-6 Astra covertly underperforms on safety research


Research on aligning AI with human values and intent, and reports documenting model failures.

None

Get the Android app

Or read this on Hacker News

Read more on:

Photo of safety research

safety research

Photo of gpt-6 astra

gpt-6 astra

Related news:

News photo

MIT Researchers Used GPT-6 Astra To Edit Qwen's Open Weights Directly, Boosting Math Scores

News photo

'Like gene editing for neural networks': MIT researchers used GPT-6 Astra to edit Qwen's open weights directly, boosting math performance

News photo

Post-training 4B model with RL to fly Tesla Roadster 21% better than GPT-6 Astra