{"slug": "tree-ising-repairing-language-models-with-bit-level-ising-edits", "title": "Tree-Ising: Repairing Language Models with Bit-Level Ising Edits", "summary": "An MIT-developed frontier LLM agent, GPT-6 Astra, repaired reasoning loops in Qwen3.5-0.8B without gradients using weight-space programming, according to work by Takehi et al., consuming 158 agent turns, 89.7 million tokens and 50 forward passes. The technique, described as Tree-Ising bit-level editing, lets an agent read a network's weights and write a repair directly.", "body_md": "Weight-space programming, introduced in recent MIT work by Takehi et al., enables a frontier LLM agent to read a network's weights and write a repair. Their GPT-6 Astra agent corrected the reasoning loops of Qwen3.5-0.8B without gradients, using 158 agent turns, 89.7 million tokens, and 50 forward-p", "url": "https://wpnews.pro/news/tree-ising-repairing-language-models-with-bit-level-ising-edits", "canonical_source": "https://aiflash.com/news/149591/", "published_at": "2026-11-11 05:30:09+00:00", "updated_at": "2026-11-11 05:51:43.426292+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-research", "machine-learning"], "entities": ["MIT", "Takehi et al.", "GPT-6 Astra", "Qwen3.5-0.8B", "Tree-Ising"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/tree-ising-repairing-language-models-with-bit-level-ising-edits", "markdown": "https://wpnews.pro/news/tree-ising-repairing-language-models-with-bit-level-ising-edits.md", "text": "https://wpnews.pro/news/tree-ising-repairing-language-models-with-bit-level-ising-edits.txt", "jsonld": "https://wpnews.pro/news/tree-ising-repairing-language-models-with-bit-level-ising-edits.jsonld"}}