Metraipolis › archive › dc_20cc5cde407d
The Invocation Bottleneck in LLM-Based Lossless Compression: What Can Be Removed and What Must Be Traded — arXiv:2611.05871
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.05871
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-05T18:28:54.803Z
- SHA-256 of the archived text
f0a47b652166fe02a1c058f4642e439fd916a2ad3f5349a3a7ff1a69e5df385b- Stable URL of this page
https://metraipolis.com/documento/dc_20cc5cde407d
How to cite
The Invocation Bottleneck in LLM-Based Lossless Compression: What Can Be Removed and What Must Be Traded — arXiv:2611.05871 — arXiv (en), https://arxiv.org/abs/2611.05871 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_20cc5cde407d (Metraipolis archive dc_20cc5cde407d, sha256 f0a47b652166fe02a1c058f4642e439fd916a2ad3f5349a3a7ff1a69e5df385b)
BibTeX
@misc{dc20cc5cde407d,
title = {The Invocation Bottleneck in LLM-Based Lossless Compression: What Can Be Removed and What Must Be Traded — arXiv:2611.05871},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.05871},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_20cc5cde407d; sha256 f0a47b652166fe02a1c058f4642e439fd916a2ad3f5349a3a7ff1a69e5df385b},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.05871. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Large language models (LLMs) compress text far better than classical compressors, but LLM-based compressors are slow, particularly during decompression. We argue that their decoding time has two fundamentally different sources. Execution redundancy, such as per-token host transfers, CPU-side coding, and token-by-token encoding, is work that the coded output does not require and can be removed without changing a single output bit. Sequential dependency, one model call per decoded token, is inherent to lossless decoding and cannot be removed by better execution, only traded for more calls or for more bits. We remove execution redundancy with a GPU-resident pipeline whose block-parallel encoding is made lossless in half precision by batch-invariant kernels, speeding up arithmetic coding by up to 5.8x in encoding and 3.0x in decoding without changing its compression rate. We then study the two ways of trading sequential dependency, speculative dual-model coding (SDMC) and multi-token single-model coding (MSMC), characterize the compression cost of predicting several tokens from a shared prefix, and identify the workloads in which fewer model calls are worth that cost. Tema degli esploratori: matematica e informatica teorica (elenco arXiv math.IT, nuovi lavori del giorno). Categorie arXiv: Information Theory (cs.IT) ; Databases (cs.DB). Pagina del lavoro: https://arxiv.org/abs/2611.05871 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-11-05T18:18:48.966Z dalla pagina https://arxiv.org/list/math.IT/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).