{"slug": "we-let-an-ai-agent-edit-a-tesseract-model-s-weights-directly-it-beat-a-last-not", "title": "We let an AI agent edit a Tesseract model's weights directly; it beat a last-layer refit, not fine-tuning", "summary": "An AI agent using weight-space programming cut errors in Tesseract's Hebrew text recognizer, a 378K-weight model, by more than a third across three gradient-free editing runs, but still lost to conventional fine-tuning, Rivok Labs reported. The agent, Claude Opus 5.5, edited the weights directly and explained each change, including finding two start-up faults in the bidirectional LSTM, one of which swallowed numbers at the start of a Hebrew line. Rivok Labs said the result shows weight-space editing can beat a last-layer refit without a training loop, though it cautioned that larger models may behave differently.", "body_md": "e tried weight-space programming (Takehi, Teoh, Zhang & Khattab, MIT CSAIL) on a small production model: Tesseract’s Hebrew text recognizer, 378K weights.\n\nAn AI agent (Claude Opus 5.5) edited the weights directly, with no gradients, three times:\n\nThe fixes were explainable: the agent found two start-up faults in the bidirectional LSTM, one of which swallowed numbers at the start of a Hebrew line.\n\nFine-tuning won this round, but we still find the result impressive: with no gradients and no training loop, an agent cut the errors by more than a third and could explain every change it made. This was a small model, and larger ones may behave differently. As agents get better, we believe weight-space editing will become a far more valuable tool.\n\nWrite-up with the negative results: [Weight-Space Programming: An AI Agent vs Fine-Tuning — Rivok Labs](https://rivoklabs.com/research/weight-space-programming?ref=hf-forum)\n\nThe fine-tuned Hebrew models we ship: [Rivok/tesseract-heb-rivok · Hugging Face](https://huggingface.co/Rivok/tesseract-heb-rivok)\n\nHappy to answer questions about the setup.", "url": "https://wpnews.pro/news/we-let-an-ai-agent-edit-a-tesseract-model-s-weights-directly-it-beat-a-last-not", "canonical_source": "https://discuss.huggingface.co/t/we-let-an-ai-agent-edit-a-tesseract-models-weights-directly-it-beat-a-last-layer-refit-not-fine-tuning/192205#post_1", "published_at": "2026-11-20 09:45:31+00:00", "updated_at": "2026-11-20 10:14:44.227619+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "ai-agents", "ai-research"], "entities": ["Rivok Labs", "Tesseract", "Claude Opus 5.5", "MIT CSAIL", "Takehi", "Teoh", "Zhang", "Khattab"], "also_reported_by": [], "takeaway": "An agent editing a 378K-weight Tesseract Hebrew recognizer's weights directly, with no gradients, cut errors by more than a third and explained each fix, but still trailed full fine-tuning.", "recommendation": null, "alternates": {"html": "https://wpnews.pro/news/we-let-an-ai-agent-edit-a-tesseract-model-s-weights-directly-it-beat-a-last-not", "markdown": "https://wpnews.pro/news/we-let-an-ai-agent-edit-a-tesseract-model-s-weights-directly-it-beat-a-last-not.md", "text": "https://wpnews.pro/news/we-let-an-ai-agent-edit-a-tesseract-model-s-weights-directly-it-beat-a-last-not.txt", "jsonld": "https://wpnews.pro/news/we-let-an-ai-agent-edit-a-tesseract-model-s-weights-directly-it-beat-a-last-not.jsonld"}}