Claude weight edit lifts Qwen’s math score to 58.1% in follow-up study
AI-generated summary A preprint listed on arXiv on November 17 stated that Claude Opus 5.5 raised a small Qwen model’s math score to 58.1% by editing its weights. Subham Kundu added the Qwen experiment to public research notes on November 16 after testing Qwen3.5-0.8B.