# MIT CSAIL: GPT-6 Astra edits Qwen's parameters

Digest AI · Research · published 2026-11-03T15:02:00Z

Canonical: https://digestai.news/story/mit-csail-gpt-6-astra-edits-qwen-s-parameters

## Summary

Researchers at MIT CSAIL have published initial results on their project "Weight-Space Programming," which uses LLMs to read and write the parameters of other neural networks. They report that GPT-6 Astra directly edited 0.016% of the parameters of Qwen3.5-0.8B, reducing thought loops where it repeats thoughts without reaching an answer. This improvement was seen with up to 50 forward passes for practice problems, but not with a simpler model like Gemma 3 270M.

## Key points

- MIT CSAIL researchers tested GPT-6 Astra on Qwen3.5-0.8B
- GPT-6 Astra edited 0.016% of Qwen's parameters
- Improvement seen with up to 50 forward passes

## Why it matters

This research could lead to more efficient and less repetitive AI models, potentially improving the performance and reliability of large language models.

## Sources

1. [AI Morning Paper | November 4: GPT-6 Astra directly edits parameters of other AIs, Claude supports automatic connection to services](https://note.com/xmatome/n/nfc494949712a?hl=en) (note.com, 2026-11-03)

Part of the developing story: [Moonshot AI Leverages GPT-6 Astra Weight Editing](https://digestai.news/thread/moonshot-ai-releases-kimi-k3-1-pro-with-1m-token-context-and-open-weights-by) (7 stories)

## Cite

Digest AI, "MIT CSAIL: GPT-6 Astra edits Qwen's parameters", 3 November 2026, https://digestai.news/story/mit-csail-gpt-6-astra-edits-qwen-s-parameters

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/mit-csail-gpt-6-astra-edits-qwen-s-parameters.json