WaveCommitments: Private and Verifiable Speech-Agent Marketplaces

查看原文 / View Original 编辑 / Edit

Computer Science > Computation and Language arXiv:2610.26867 (cs) [Submitted on 16 Jul 2026] Title:WaveCommitments: Private and Verifiable Speech-Agent Marketplaces Authors:Oguzhan Baser, Kaan Kale, Sriram Vishwanath, Sandeep Chinchali View a PDF of the paper titled WaveCommitments: Private and Verifiable Speech-Agent Marketplaces, by Oguzhan Baser and 3 other authors View PDF HTML (experimental)

计算机科学 > 计算与语言 arXiv:2610.26867 (cs) [提交于 2026 年 7 月 16 日] 标题：WaveCommitments：私有且可验证的语音代理市场作者：Oguzhan Baser, Kaan Kale, Sriram Vishwanath, Sandeep Chinchali 查看标题为“WaveCommitments：私有且可验证的语音代理市场”的论文 PDF，作者为 Oguzhan Baser 及其他 3 位作者查看 PDF HTML（实验性）

Abstract:Modern speech systems are evolving into ecosystems of autonomous speech agents that speak together, negotiate, and trade assets such as embeddings, spectrograms, or raw data. In such markets, sellers cannot reveal assets for evaluation due to privacy constraints, and buyers cannot price what they cannot test.

摘要：现代语音系统正在演变为自主语音代理的生态系统，这些代理能够协同对话、协商并交易嵌入、频谱图或原始数据等资产。在此类市场中，由于隐私限制，卖方无法披露资产以供评估，而买方也无法为无法测试的资产定价。

Thus, we open-source WaveCommitments, an end-to-end design that pairs confidential asset valuation via Data Shapley inside a trusted execution environment with multivariate polynomial commitments. It enables selective openings and tamper-evident provenance to prove that a delivered asset matches the valuation while hiding everything else.

因此，我们开源了 WaveCommitments，这是一种端到端的设计，它将受信任执行环境内的 Data Shapley 机密资产评估与多元多项式承诺相结合。它支持选择性披露和防篡改溯源，以证明交付的资产与评估结果相符，同时隐藏所有其他信息。

Our extensive experiments across three known datasets and various tasks with state-of-the-art encoders, including Whisper, reveal that we speed up the best benchmarks by up to 20x while reducing their proof sizes by up to 39x. We also systematically study how a speaker’s value shifts across tasks and encoders, revealing agent-level market prices.

我们在三个已知数据集上，利用包括 Whisper 在内的最先进编码器进行了广泛的实验，结果表明，我们的方法将最佳基准测试的速度提高了 20 倍，同时将证明大小减少了 39 倍。我们还系统地研究了说话者的价值如何在不同任务和编码器之间发生变化，从而揭示了代理级别的市场价格。