HN 日本語サマリー

← 一覧へ戻る
AI・機械学習

Show HN: 46GB GPU上で600万トークンの移動ウィンドウを実現

Show HN: A 6M-token movable window on a single 46GB GPU (arxiv.org)

6 pointsby Wetime7 コメント

要約

本論文では、言語モデルの改善に再学習ではなく、モデルを固定し検証済みソリューションの永続的なメモリを成長させるアプローチを提案しています。これにより、一度解決された問題ファミリーは、生成トークンゼロでビット単位で正確かつ決定論的に回答できるようになります。このシステムは、単一の46GB GPU上で600万トークンという大規模な移動ウィンドウを可能にし、既存の最先端モデルを凌駕する性能を示しています。

全文翻訳

Computer Science > Computation and Language

arXiv:2607.23806 (cs)

[Submitted on 26 Jul 2026]

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

Authors:[Sietse Schelpe](https://arxiv.org/search/cs?searchtype=author&query=Schelpe,+S) (Corbenic AI)

View a PDF of the paper titled A Frozen 12B Beats Frontier Models on Verified Work: 100% Accuracy, 0 Tokens, Bit-Exact, Forever, by Sietse Schelpe (Corbenic AI)

[View PDF](https://arxiv.org/pdf/2607.23806)

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. 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: [this https URL](https://corbenic-galahad-bench.hf.space/)

Comments:Industry experience report. 14 pages, 8 figures. Public testbench: [this https URL](https://corbenic-galahad-bench.hf.space/); companion repository with SHA-256 provenance manifest: [this https URL](https://github.com/corbenicai/galahad-bench)
Subjects:Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR); Machine Learning (cs.LG); Performance (cs.PF)
Cite as:[arXiv:2607.23806](https://arxiv.org/abs/2607.23806) [cs.CL]
(or [arXiv:2607.23806v1](https://arxiv.org/abs/2607.23806v1) [cs.CL] for this version)
[https://doi.org/10.48550/arXiv.2607.23806](https://doi.org/10.48550/arXiv.2607.23806) arXiv-issued DOI via DataCite

Submission history

From: Sietse Schelpe [[view email](https://arxiv.org/show-email/d8f3a1c9/2607.23806)]
[v1] Sun, 26 Jul 2026 19:08:33 UTC (1,421 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled A Frozen 12B Beats Frontier Models on Verified Work: 100% Accuracy, 0 Tokens, Bit-Exact, Forever, by Sietse Schelpe (Corbenic AI)

  • [[View PDF](https://arxiv.org/pdf/2607.23806)]

[[画像: license icon]] [[view license](http://creativecommons.org/licenses/by/4.0/)]

Additional Features
  • [[Audio Summary](https://arxiv.org/audio/2607.23806)]
Current browse context:

cs.CL

[< prev](https://arxiv.org/prevnext?id=2607.23806&function=prev&context=cs.CL)   |   [[next >](https://arxiv.org/prevnext?id=2607.23806&function=next&context=cs.CL)]

[new](https://arxiv.org/list/cs.CL/new) | [[recent](https://arxiv.org/list/cs.CL/recent) | [[2026-07](https://arxiv.org/list/cs.CL/2026-07)]

Change to browse by:

[cs](https://arxiv.org/abs/2607.23806?context=cs)
[cs.AI](https://arxiv.org/abs/2607.23806?context=cs.AI)
[cs.IR](https://arxiv.org/abs/2607.23806?context=cs.IR)
[cs.LG](https://arxiv.org/abs/2607.23806?context=cs.LG)
[cs.PF](https://arxiv.org/abs/2607.23806?context=cs.PF)

References & Citations
  • [[SciX](https://scixplorer.org/abs/arXiv:2607.23806)]
  • [[Google Scholar](https://scholar.google.com/scholar_lookup?arxiv_id=2607.23806)]
  • [[Semantic Scholar](https://api.semanticscholar.org/arXiv:2607.23806)]

Loading...

BibTeX formatted citation

loading...

Data provided by:

Bookmark

[[画像: BibSonomy]](http://www.bibsonomy.org/BibtexHandler?requTask=upload&url=https://arxiv.org/abs/2607.23806&description=A%20Frozen%2012B%20Beats%20Frontier%20Models%20on%20Verified%20Work:%20100%%20Accuracy,%200%20Tokens,%20Bit-Exact,%20Forever)[[画像: Reddit]](https://reddit.com/submit?url=https://arxiv.org/abs/2607.23806&title=A%20Frozen%2012B%20Beats%20Frontier%20Models%20on%20Verified%20Work:%20100%%20Accuracy,%200%20Tokens,%20Bit-Exact,%20Forever)

Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle

Bibliographic Explorer ([What is the Explorer?](https://info.arxiv.org/labs/showcase.html#arxiv-bibliographic-explorer))

Connected Papers Toggle

Connected Papers ([What is Connected Papers?](https://www.connectedpapers.com/about))

Litmaps Toggle

Litmaps ([What is Litmaps?](https://www.litmaps.co/))

scite.ai Toggle

scite Smart Citations ([What are Smart Citations?](https://www.scite.ai/))

Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle

alphaXiv ([What is alphaXiv?](https://alphaxiv.org/))

Links to Code Toggle

CatalyzeX Code Finder for Papers ([What is CatalyzeX?](https://www.catalyzex.com/))

DagsHub Toggle

DagsHub ([What is DagsHub?](https://dagshub.com/))

GotitPub Toggle

Gotit.pub ([What is GotitPub?](http://gotit.pub/faq))

Huggingface Toggle

Hugging Face ([What is Huggingface?](https://huggingface.co/huggingface))

ScienceCast Toggle

ScienceCast ([What is ScienceCast?](https://sciencecast.org/welcome))

Demos

Demos

Replicate Toggle

Replicate ([What is Replicate?](https://replicate.com/docs/arxiv/about))

Spaces Toggle

Hugging Face Spaces ([What is Spaces?](https://huggingface.co/docs/hub/spaces))

Spaces Toggle

TXYZ.AI ([What is TXYZ.AI?](https://txyz.ai/))

Related Papers

Recommenders and Search Tools

Link to Influence Flower

Influence Flower ([What are Influence Flowers?](https://influencemap.cmlab.dev/))

Core recommender toggle

CORE Recommender ([What is CORE?](https://core.ac.uk/services/recommender))

  • Author
  • Venue
  • Institution
  • Topic

About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2607.23806) | Disable MathJax ([[What is MathJax?](https://info.arxiv.org/help/mathjax.html))]