aiminute. ← All AI news
Research AI Minute Newsroom 2026-11-21

Researchers say an open AI model improved by editing its own weights.

Researchers say an open AI model improved by editing its own weights.

A preprint posted this week describes an open Qwen model from Alibaba changing its own weights without any training. Its edits raised accuracy on unseen problems by nearly seven points. Smaller models from the same family failed or broke themselves, and the work is not yet peer reviewed.

Why it mattersModels normally change only when engineers retrain them with large amounts of computing. The authors call the result a step toward AI that improves itself, a prospect safety researchers watch closely.

✓ Verified · 2 sources

WhatsApp X Telegram
Read in the app — free, in 9 languages

Related stories

OpenAI says its new models block nearly all reasoning-theft attacks.
2026-11-22
Lean's own community says the AI-assisted Riemann proof is correct.
2026-11-22
A US watchdog found that no AI therapy chatbot has FDA authorization.
2026-11-21
Outside mathematicians say the AI Riemann proof passed first checks.
2026-11-21
A test with two million bait prompts caught AI services leaking data.
2026-11-21