arXiv：cs.CL· Mengzhe Geng·· 4 小时前AI 评分37
PACT-SLM 合约测试：语音智能体何时有足够证据行动？
When Does a Spoken Agent Have Enough Evidence to Act? The PACT-SLM Contract Test
AI 导读
研究者提出 PACT-SLM 合约测试，为流式语音智能体分配首个有效动作时间，同时评估动作身份与时机。主测试用四个留出语义家族的 80 组配对对比，在干净与 15 dB 噪声下产生 1,600 次前缀预测；WavLM Base Plus 的 onset 后语义标签池化准确率为 26.03%，早于 onset 前缀中 18.99% 暴露动作，完整轨迹精确预测率 5.94%。
正文
Abstract:Streaming spoken agents may produce the correct final action after acting too early. Final-turn scores do not reveal whether each observed speech prefix supports an exposed action. We introduce the Partial Speech Action Contract for Turn Taking in Speech Language Models (PACT-SLM), a controlled test that assigns the first valid action time and evaluates both action identity and timing. In the primary test, 80 paired contrast groups from four held-out semantic families yield 1,600 prefix predictions across clean and 15 dB noise renderings. Using source-utterance semantic targets rather than counterbalanced branch codes, WavLM Base Plus reaches 26.03% pooled post-onset semantic-label accuracy (95% group-bootstrap interval [22.14%, 29.68%]), exposes an action on 18.99% of pre-onset prefixes, and predicts 5.94% of complete trajectories exactly. Its pooled label score is at the 96th percentile of 100 within-prefix label permutations, below the 97.5th-percentile reference (26.73%). It exceeds matched text, scalar-acoustic, and shuffled-representation probes in pooled post-onset label accuracy. Elapsed time is more onset-exact (36.25% versus 23.13%) but less accurate about action identity (9.92% versus 26.03%). These results motivate separate measurement of action identity and onset timing in partial-speech evaluations.
| Subjects: | Sound (cs.SD); Computation and Language (cs.CL); Audio and Speech Processing (eess.AS) |
| Cite as: | arXiv:2609.38232 [cs.SD] |
| (or arXiv:2609.38232v2 [cs.SD] for this version) | |
| https://doi.org/10.48550/arXiv.2609.38232 arXiv-issued DOI via DataCite |
Submission history
From: Mengzhe Geng [view email]
[v1]
Mon, 28 Sep 2026 17:11:17 UTC (164 KB)
[v2]
Wed, 7 Oct 2026 11:02:46 UTC (161 KB)
来源：arXiv：cs.CL · arxiv.org