A Frozen 12B Beats Frontier Models on Verified Work: 100% Accuracy, 0 Tokens, Bit-Exact, Forever
- View PDF Abstract: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 generation tokens, bit-exact, deterministically.
Unverified
- View PDF Abstract: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 generation tokens, bit-exact, deterministically.
Sources: Arxiv