Metraipolis › archive › dc_7a6984a03343
PerfReasoning: How Well Do LLMs Reason on Hardware Performance? — arXiv:2609.04476
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.04476
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T22:18:41.295Z
- SHA-256 of the archived text
d14f5b45d6587bc577fddc4728c790b022d61bf51350657242d1daa7b15bac0c- Stable URL of this page
https://metraipolis.com/documento/dc_7a6984a03343
How to cite
PerfReasoning: How Well Do LLMs Reason on Hardware Performance? — arXiv:2609.04476 — arXiv (en), https://arxiv.org/abs/2609.04476 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_7a6984a03343 (Metraipolis archive dc_7a6984a03343, sha256 d14f5b45d6587bc577fddc4728c790b022d61bf51350657242d1daa7b15bac0c)
BibTeX
@misc{dc7a6984a03343,
title = {PerfReasoning: How Well Do LLMs Reason on Hardware Performance? — arXiv:2609.04476},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.04476},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_7a6984a03343; sha256 d14f5b45d6587bc577fddc4728c790b022d61bf51350657242d1daa7b15bac0c},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.04476. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Performance modeling is central to hardware design and software optimization, yet constructing these models requires structured reasoning about computation, data reuse, storage, and movement. We introduce PerfReasoning, a benchmark that evaluates LLMs both as direct performance reasoners and as generators of analytical performance-model code. Given workload, architecture, and mapping specifications, models compare mappings and predict off-chip traffic and buffer requirements. The strongest closed-source models exceed 90% on reasoning-based Q&A, and the best open-weight model reaches 82.4%. However, model construction is substantially harder: while GPT-5.6 Sol exceeds 80% pass rate, all other model configurations average below 45% and vary markedly across runs. Task-specific RL raises a 4B model's mapping-reasoning accuracy by 15.7 points, whereas feedback-free multi-round self-revision prompting is not reliably effective. PerfReasoning exposes the gap between plausible architectural reasoning and reliable performance-model construction. We will publicly release the benchmark to support reproducible evaluation and track future progress. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-07 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI) ; Performance (cs.PF). Pagina del lavoro: https://arxiv.org/abs/2609.04476 — 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 2026-10-16T22:18:11.347Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-07?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).