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

arXiv: https://arxiv.org/abs/2607.23806 · HTML: https://mlsanity.com/paper/2607.23806/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 18 of its 29 references that are in the map.
- First posted on arXiv 26 Jul 2026.

## Builds on (most cited first, 18 of 18)

- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [Self-Consistency Improves Chain of Thought Reasoning in Language Models](https://mlsanity.com/paper/2203.11171/) (2022)
- [SGLang: Efficient Execution of Structured Language Model Programs](https://mlsanity.com/paper/2312.07104/) (2023)
- [LeanDojo: Theorem Proving with Retrieval-Augmented Language Models](https://mlsanity.com/paper/2306.15626/) (2023)
- [CodeT: Code Generation with Generated Tests](https://mlsanity.com/paper/2207.10397/) (2022)
- [Large Language Models as Tool Makers](https://mlsanity.com/paper/2305.17126/) (2023)
- [Prompt Cache: Modular Attention Reuse for Low-Latency Inference](https://mlsanity.com/paper/2311.04934/) (2023)
- [Gemma 4 Technical Report](https://mlsanity.com/paper/2607.02770/) (2026)
- [CacheBlend: Fast Large Language Model Serving for RAG with Cached Knowledge Fusion](https://mlsanity.com/paper/2405.16444/) (2024)
- [Memp: Exploring Agent Procedural Memory](https://mlsanity.com/paper/2508.06433/) (2025)
- [TroVE: Inducing Verifiable and Efficient Toolboxes for Solving Programmatic Tasks](https://mlsanity.com/paper/2401.12869/) (2024)
- [Agentic Plan Caching: Test-Time Memory for Fast and Cost-Efficient LLM Agents](https://mlsanity.com/paper/2506.14852/) (2025)
- [From Lossy to Verified: A Provenance-Aware Tiered Memory for Agents](https://mlsanity.com/paper/2602.17913/) (2026)
- [VeriCache: Turning Lossy KV Cache into Lossless LLM Inference](https://mlsanity.com/paper/2605.17613/) (2026)
- [Byte-Exact Deduplication in Retrieval-Augmented Generation: A Three-Regime Empirical Analysis Across Public Benchmarks](https://mlsanity.com/paper/2605.09611/) (2026)
- [Portable Agent Memory: A Protocol for Cryptographically-Verified Memory Transfer Across Heterogeneous AI Agents](https://mlsanity.com/paper/2605.11032/) (2026)
- [Merlin: Deterministic Byte-Exact Deduplication for Lossless Context Optimization in Large Language Model Inference](https://mlsanity.com/paper/2605.09990/) (2026)
- [Trained Persistent Memory for Frozen Decoder-Only LLMs](https://mlsanity.com/paper/2603.22329/) (2026)

## Built on by (most cited first, 2 of 2)

- [Working Around the Compute Ceiling: Byte-Exact Memory in Galahad Makes LLM Reading a One-Time Cost LLM Reading a One-Time Cost](https://mlsanity.com/paper/2609.39358/) (2026)
- [Real Long-Term Memory for AI: A 50-Million-Token Window That Is Faster and Cheaper Than Recompute](https://mlsanity.com/paper/2610.10845/) (2026)

## Abstract (arXiv)

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. Across 180 fresh instances spanning nine problem families, four architectures from four vendors - dense and mixture-of-experts - each score 180/180 at zero generation tokens per answer: execution-bound capability decoupled from parameter scaling. A negative control attributes the capability fully to the memory: emptied, it solves nothing. The same verify-before-store contract holds for open-ended reasoning: 88/88 consistency-gated acceptances across all four models, machine-checked formal proof, and reasoning-method transfer at 77/80. Memory selection takes 1.4 microseconds; a full reuse completes in 6-23 ms at 36 mWh. Approximate similarity retrieval selects the wrong item 94.3% of the time on a 4,500-item verified store where exact addressing makes zero errors. The store also serves as working context at a scale no shipped engine matches: a 6,000,000-token movable window on a single 46 GB GPU at flat memory, where vLLM stops at 30,399 tokens and SGLang silently truncates past 32,000. On published benchmarks, frontier models remain far ahead of any 12B at raw from-scratch reasoning; on everything this system has solved and verified, the comparison inverts: a frontier API call pays a fresh generation pass on every query, forever, while verified reuse costs zero tokens and returns the identical bits every time. A public testbench with free, rate-limited access accompanies this report: https://corbenic-galahad-bench.hf.space