**A Frozen 12B Beats Frontier Models on Verified Work: 100% Accuracy, 0 Tokens,

**A Frozen 12B Beats Frontier Models on Verified Work: 100% Accuracy, 0 Tokens, Bit-Exact, Forever**

> Improving a language model today means retraining it: enormous compute, a new opaque model each cycle, non-deterministic output. We take the opposite path: the model stays frozen, and a persistent memory of verified solutions grows beside it. Once a problem family is solved and has passed an independent verification step that never consults the answer key, every new instance of that family is answered at zero generat...

**来源信息**
- **来源**：HuggingFace Daily Papers（社区热门论文）
- **分类**：ai-products
- **发布时间**：2026-07-28 08:00（北京时间）
- **原文**：[打开原文](https://huggingface.co/papers/2607.23806)