Metraipolis › archive › dc_61a306da7a06

Reasoning-Aware Compression: Identifying and Protecting Vulnerable Reasoning Circuits for Energy-Efficient LLM Deployment — arXiv:2609.05512

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.05512
License (as archived)
CC0-1.0
Archived
2026-10-16T12:49:41.972Z
SHA-256 of the archived text
a0773bafbde53f3de6b39cb23b03eff193d55fc2a2dc12c86ece02a31d2da2c6
Stable URL of this page
https://metraipolis.com/documento/dc_61a306da7a06

How to cite

Reasoning-Aware Compression: Identifying and Protecting Vulnerable Reasoning Circuits for Energy-Efficient LLM Deployment — arXiv:2609.05512 — arXiv (en), https://arxiv.org/abs/2609.05512 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_61a306da7a06 (Metraipolis archive dc_61a306da7a06, sha256 a0773bafbde53f3de6b39cb23b03eff193d55fc2a2dc12c86ece02a31d2da2c6)
BibTeX
@misc{dc61a306da7a06,
  title = {Reasoning-Aware Compression: Identifying and Protecting Vulnerable Reasoning Circuits for Energy-Efficient LLM Deployment — arXiv:2609.05512},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.05512},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_61a306da7a06; sha256 a0773bafbde53f3de6b39cb23b03eff193d55fc2a2dc12c86ece02a31d2da2c6},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.05512. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Large Reasoning Models (LRMs) impose substantial energy costs during deployment, yet current compression methods apply uniform quantization across all components, risking damage to critical reasoning circuits. We present a reasoning-aware compression framework that benchmarks quantization conditions across five reasoning benchmarks, GSM8K, FOLIO, MATH-500, ProofWriter, and MuSiQue, with hardware-level GPU energy measurement; profiles per-module INT4 vulnerability across all 196-224 (layer, projection) pairs via a perturbation sweep on a held-out calibration split, then selectively restores the most sensitive circuits to FP16. Three findings emerge. First, INT4 quantization can increase energy by extending reasoning chains; a 25% power reduction becomes a net energy increase on GSM8K. Second, vulnerability is task-dependent: attention projections are more critical for mathematical reasoning, and sensitivity patterns differ by architecture in logical inference. Third, selective compression achieves Pareto-optimal points inaccessible to uniform methods: R1-Qwen-7B Top-10% on ProofWriter gains +12 pp over FP16 at -9.7% energy, validated on held-out data across five reasoning benchmarks.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-09 in cs.CL).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Computation and Language (cs.CL); Performance (cs.PF).
Pagina del lavoro: https://arxiv.org/abs/2609.05512 — 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 2026-10-16T12:30:25.229Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-09?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).