arXiv cs.AI· Oguzhan Baser, Elahe Sadeghi, Eric Wang, Nico Vergauwen, Sam Kazemian, Hong Kang, Sandeep P. Chinchali, Sriram Vishwanath·· 3 小时前AI 评分44
TensorCommitments：面向大语言模型的轻量级可验证推理方案
TensorCommitments: A Lightweight Verifiable Inference for Large Language Models
AI 导读
研究者提出 TensorCommitments（TC），一种基于多元 Terkle Trees 的张量原生推理证明方案，让用户无需重跑模型即可验证云端 LLM 推理是否被篡改。在 LLaMA2 上，TC 仅增加 0.97% 证明方和 0.12% 验证方时间开销，对定制化攻击的鲁棒性比此前依赖验证方 GPU 的最优方案提升最多 48%。
正文 · 原文
Abstract:Most large language models (LLMs) run on external clouds: users send a prompt, pay for inference, and must trust that the remote GPU executes the LLM without any adversarial tampering. We critically ask how to achieve verifiable LLM inference, where a prover (the service) must convince a verifier (the client) that an inference was run correctly without rerunning the LLM. Existing cryptographic works are too slow at the LLM scale, while non-cryptographic ones require a strong verifier GPU. We propose TensorCommitments (TCs), a tensor-native proof-of-inference scheme. TC binds the LLM inference to a commitment, an irreversible tag that breaks under tampering, organized in our multivariate Terkle Trees. For LLaMA2, TC adds only 0.97% prover and 0.12% verifier time over inference while improving robustness to tailored attacks by up to 48\% over the best prior work requiring a verifier GPU.
| Subjects: | Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2602.12630 [cs.CR] |
| (or arXiv:2602.12630v3 [cs.CR] for this version) | |
| https://doi.org/10.48550/arXiv.2602.12630 arXiv-issued DOI via DataCite |
Submission history
From: Oguzhan Baser [view email]
[v1]
Fri, 13 Feb 2026 05:23:31 UTC (4,832 KB)
[v2]
Thu, 1 Oct 2026 09:29:49 UTC (4,832 KB)
[v3]
Fri, 23 Oct 2026 00:09:59 UTC (4,819 KB)
来源：arXiv cs.AI · arxiv.org