MIT researchers show GPT-6 Astra can rewrite other models' weights directly
The finding suggests frontier models may be able to reprogram peer models without conventional training.
Read at The Decoder โFull coverage ยท 12 outlets
- The Decoder ยท lead
- VentureBeat'Like gene editing for neural networks': MIT researchers used GPT-6 Astra to edit Qwen's open weights directly, boosting math performance
- Startup FortuneMIT Researchers Show GPT-6 Astra Can Rewrite Another AI's Brain Without Training It
- The RegisterForget retraining - AI can now tweak another AI's weights
- The Rundown AIMIT shows AI improving AI, no training required
- BigGo FinanceNo Training, No Gradients: MIT Lets GPT-6 Astra Rewrite Qwen Weights Directly, Lifting Math Score From 7.9% to 48.1%
- news.lavx.huMIT researchers show AI can edit another model's weights directly
- R&D WorldMIT team says GPT-6 Astra can improve other AI models by editing their weights. Critics say its biggest result relied on standard ML engineering.
- Analytics India MagazineMIT Researchers Show LLMs Can Read and Edit Neural Network Weights Without Retraining
- Pasquale PillitteriGPT-6 Astra Fixes a Flaw in Qwen's Weights, MATH-500 Jumps From 7.9% to 48.1%
- OfficeChaiAbility For Weight-Space Programming Has Appeared Only In Latest Models, Says Google Researcher Alex Krentsel Who Reproduced Results
- BlockchainMIT CSAIL: GPT-6 Astra rewrites weights to boost Qwen math