arXiv AI July 17, 2026 By Sietse Schelpe

Smarter and Cheaper at Once: Byte-Exact KV-Cache Grafting Turns a Frozen Small Model into a Verified-Knowledge Flywheel

arXiv:2607. 14431v1 Announce Type: cross Abstract: We report a way to make a frozen small language model both more capable and dramatically cheaper at once, without changing any weights.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.

llms

arXiv AI

Jul 28

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

arXiv:2607. 23806v1 Announce Type: cross Abstract: Improving a language model today means retraining it: enormous compute, a new opaque model each cycle, non-deterministic output.

By Sietse Schelpe (Corbenic AI)

llms benchmarks

arXiv Machine Learning

Jul 31

Stage-Replay Divergence Follows the KV Cache: Fixed-Prefix Precision Controls and Bidirectional Cache Transplantation

arXiv:2607. 28495v1 Announce Type: new Abstract: Stage-replay diagnostics reconstruct intermediate token prefixes and treat fresh-prefill continuation as continuation from the decoder state that originally reached the prefix.

By Alexander Boesgaard Lorup

efficiency

arXiv AI

Jun 11

Compiler-First State Space Duality and Portable $O(1)$ Autoregressive Caching for Inference

arXiv:2603. 09555v2 Announce Type: replace-cross Abstract: High-throughput Mamba-2 inference is usually tied to fused CUDA and Triton kernels, limiting portability across accelerator backends.

By Cosmo Santoni, Anmol Thapar

rag efficiency

arXiv Machine Learning

4d ago

Behavioral Capacity Certificates for Quantized Language Models

arXiv:2609.37887v1 Announce Type: new Abstract: Activation and key-value cache precision change what a quantized language model computes without altering its stored weights. Direct weight-code bounds...

By Arian Eamaz, Mojtaba Soltanalian

llms efficiency

arXiv Machine Learning

Aug 11

RotaryQuant: Fitting 120B MoE Models on Consumer Hardware via Fused Compressed-Space Attention

arXiv:2608. 08081v1 Announce Type: cross Abstract: Large mixture-of-experts (MoE) language models with 26--120 billion parameters exceed the memory capacity of consumer devices through three simultaneous pressures: resident weight matrices, key-value (KV) cache state that grows linearly with context, and dozens of expert sublayers that must be paged on demand.

By Anthony. Lui, Mohamed. Elsaied, N. P. Savani

llms efficiency

arXiv AI

Jul 17

A Modern Multimodal Assistant on a 6 GB 2011 GPU: Stage-Validated, All-GPU CUDA Inference for Fermi

arXiv:2607. 14568v1 Announce Type: cross Abstract: A companion study ran a 35B mixture-of-experts model on a 2011 NVIDIA Tesla C2075 (Fermi, sm_20, 6GB) as a GPU-prefill/CPU-decode hybrid, because the 4-bit model did not fit in device memory (arXiv:2606.

By A. C. Opus, J. Q. Lu

rag multimodal benchmarks

One email a morning, machine-written

Related stories

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

Stage-Replay Divergence Follows the KV Cache: Fixed-Prefix Precision Controls and Bidirectional Cache Transplantation

Compiler-First State Space Duality and Portable $O(1)$ Autoregressive Caching for Inference

Behavioral Capacity Certificates for Quantized Language Models

RotaryQuant: Fitting 120B MoE Models on Consumer Hardware via Fused Compressed-Space Attention

A Modern Multimodal Assistant on a 6 GB 2011 GPU: Stage-Validated, All-GPU CUDA Inference for Fermi