Daily Papers — 2026-07-30

9 papers on audio, speech, music, and acoustics.

1. RIPPLE: Generating Multi-Channel Phase, Not Recovering It

Authors: Jaehyuk Lee, Yeajin Lee, Dayeon Shin, Donghun Lee

Categories: cs.LG, cs.SD Score: 7.79/10 (Obj:9 Id:8 Ind:8 Comp:8 Eff:8 Nov:8 Rep:6)

Strength: 将 Griffin-Lim 从相位估计器重构为相位先验，配合解耦 rectified flow 和 IPD loss，在 FOA 空间音频和地震跨台站翻译两个不相关物理域上，将 S波偏振误差从随机水平 57.3° 降至 33.8°，并通过 Table 5 oracle 实验严格证明了逐通道相位恢复的天花板。
Weakness: 代码和 checkpoint 未开源，σprior 和训练时 K 的敏感性未测试，且方法的适用范围限于源-目标共享通道间结构的翻译任务，无法处理 text-to-spatial 等无源信号场景。
Full review: Claude Code 全文七公理审稿

Generative models synthesize magnitude spectra with high fidelity, while phase is delegated to a recovery module—Griffin–Lim, a vocoder, or a latent decoder—applied independently to each channel. For multi-channel waveforms this delegation is costly: the physical content of spatial audio and three-component seismograms lives in the phase relationships between channels, precisely what channel-independent recovery cannot produce. The cost is also invisible, since the magnitude-based metrics common to both fields barely move when inter-channel phase coherence collapses—so a pipeline can discard the physical information in its output while still scoring well. We argue that phase should be generated, not recovered, and present RIPPLE (Rectified Inter-channel Phase with Prior-based LEarning), which reinterprets Griffin–Lim as a phase prior rather than a final estimator: initialized from the source phase, this prior carries the inter-channel structure to be preserved, and a rectified flow refines it toward the target under an explicit inter-channel phase loss. Tested on first-order ambisonics environment transfer and seismic cross-station translation—two physically unrelated domains—RIPPLE outperforms recovery-based pipelines on the coherence metrics that downstream analyses consume. The seismic case is decisive: across architecturally distinct generators, per-channel recovery leaves S-wave polarization error near the $57.3^\circ$ random expectation, whereas learned phase reduces it to $33.8^\circ$.

Authors: Xilin Jiang, Riki Shimizu, Sukru Samet Dindar, Junkai Wu, Zhongweiyang Xu et al.

Categories: cs.SD, cs.CL, cs.MM, eess.AS Score: 6.32/10 (Obj:8 Id:5 Ind:4 Comp:6 Eff:8 Nov:5 Rep:8)

Strength: 在真实且重要的多说话人嘈杂语音对话新任务上，通过三 action token + SFT+GRPO 实现 macro F1 0.93（vs 最强 baseline 0.50），跨 28 个环境泛化稳健，代码和 pipeline 开源。
Weakness: GRPO 仅贡献 +2.5 pp 提升（标题强调但非主体贡献），listen/ignore 在评测中被合并削弱三分设计必要性，训练+评测数据同源（均由 Gemini 生成）构成循环依赖且无人类校验，缺少最简级联 baseline。
Full review: Claude Code 全文七公理审稿

Spoken dialog systems are typically designed for clean, dyadic interactions in which a single user and an assistant take turns speaking. Real-world social conversations, however, are often more ambiguous: multiple speakers may participate in the same conversation amid irrelevant speech and background noise. Each utterance may be directed to the assistant, addressed to another speaker, or completely irrelevant. In such settings, the assistant must decide not only what to say, but also whether to speak at all. In this paper, we introduce Cocktail-Talker, a speech LLM framework for multi-speaker spoken dialog modeling in noisy social environments. We model the assistant’s behavior with three action tokens: < respond >, < listen >, and < ignore >, placed before a response or silence. Cocktail-Talker is trained via supervised finetuning and reinforcement learning to generate the appropriate action token and, only in < respond > mode, a speech response. To prepare the training data, we develop Cocktail-DialogGen, an LLM-based data pipeline that simulates realistic multi-speaker dialogs with speaker roles across diverse social settings. Together, these components take a step toward spoken dialog systems that interact more naturally and selectively in complex social environments.

3. CrowdioSet and PaRIRset: Two Datasets Towards Live Music Source Separation

Authors: Enric Gusó, Xavier Serra

Categories: cs.SD, eess.AS | Accepted to ISMIR26. See : https://enricguso.github.io/crowdioset_parirset Score: 6.30/10 (Obj:9 Id:6 Ind:6 Comp:5 Eff:5 Nov:6 Rep:9)

Strength: 首次为现场音乐源分离提供两个公开数据集（4800 真实观众噪声 + 40 个专业音乐厅 PA 系统 RIR），主观听感测试 26 人中 105/130 和 119/130 显著偏好本文模型，且噪声增强在 clean 数据上几乎无性能损失（vocals 9.97 vs 10.05 dB）。
Weakness: 方法是 WHAMR! (2019) 从语音到音乐的直接迁移，绝对 SDR 偏低（noisyrev vocals ~3 dB，audience isolation 1-3 dB），PaRIRset 仅 1/4 增强策略达统计显著（p=0.027），缺少 HTDemucs/BSRNN 等强 baseline，四种配置使用不同学习率引入混淆。
Full review: Claude Code 全文七公理审稿

Most Music Source Separation (MSS) models do not generalize well to live music recordings because they are trained on studio recordings alone, disregarding the venue acoustics, the speaker system’s response and audience noise. We propose to bridge this gap by providing and training a model on two novel datasets. First, we present CrowdioSet: a noise dataset comprising 4800 real ambience tracks from Freesound and synthetic sing-alongs for the vocals in MUSDB18 and MOISESDB datasets, generated from zero-shot singing voice conversions. CrowdioSet enables effective audio denoising for live recordings, resulting in superior separation both in objective and subjective evaluations. Second, we introduce PaRIRset, a stereo impulse response dataset captured across 40 professional concert venues using a microphone array. Our results show that adding PaRIRset RIRs increases the performance of a MSS model compared to using real RIRs from Speech Enhancement tasks alone. We make the examples, code, model weights, PaRIRset, and CrowdioSet freely available to the public.

4. VocalRender: Score-Native Singing Voice Synthesis for Real-World Composition

Authors: Yukun Chen, Tianrui Wang, Zhaoxi Mu, Xinyu Yang, EngSiong Chng

Categories: cs.SD, cs.AI Score: 6.05/10 (Obj:8 Id:5 Ind:5 Comp:6 Eff:8 Nov:5 Rep:8)

Strength: Score-native SVS 通过 ARDM 在连续 VAE latent 上 patch-wise 自回归生成，消除显式 duration predictor，N-CMOS 比最强基线 SoulX-Singer 高 0.42，WER 4.44 接近 GT 3.81，OOD 上基线崩溃而本方法稳定。
Weakness: 关键架构变量（ARDM vs NAR-DiT 同 backbone 同数据）未消融，interleaving 消融对旋律指标几乎无效甚至 WER 变好，WER 评测与训练数据 transcription 共用 Qwen3-ASR 形成评测闭环。
Full review: Claude Code 全文七公理审稿

Existing singing voice synthesis systems often require predefined durations, explicit duration prediction, or time-aligned acoustic guidance, which limits their compatibility with practical composition workflows. We propose VocalRender, a score-native system that directly synthesizes singing from lyrics, pitches, symbolic note values, and tempo. It uses an interleaved lyric–note representation and an autoregressive diffusion model to generate continuous acoustic latents while predicting the output length, eliminating the need for explicit duration prediction. Trained on a 2,300-hour singing dataset, VocalRender achieves strong intelligibility, strong melody control, and high speaker similarity across both in-domain and out-of-domain benchmarks. Notably, it outperforms the strongest baseline by $0.42$ points in naturalness CMOS, demonstrating the effectiveness of our proposed score-native architecture.

5. Teffic-Audio: Tell Fact from Fiction

Authors: Wan Lin, Li Wang, Jindong Wang, Kunyu Feng, Zhizheng Wu

Categories: cs.SD, cs.AI | 16 pages, 1 figure, 7 tables. Technical report. Project page: https://tefficlabs.com/teffic-audio Score: 5.47/10 (Obj:8 Id:6 Ind:5 Comp:5 Eff:8 Nov:3 Rep:3)

Strength: 标准架构（w2v-BERT 2.0 + MHASP）配合精心设计的训练配方，在 Speech-DF-Arena 14 个测试集上取得 pooled EER 1.454%，超越所有公开系统（含 3B 参数的 Resemble-Detect-3B），且仅用 590M 参数，消融实验系统验证了 balanced sampling、supplementary bonafide、diverse augmentation 的各自贡献。
Weakness: 系统标记为 Proprietary、无代码/模型/checkpoint 公开、训练超参数（epoch/lr/batch size/优化器）全部未报告，且所有组件（编码器/池化/增强/采样策略）均为已有方法的直接组合，缺乏机制层面的新颖性解释。
Full review: Claude Code 全文七公理审稿

Speech deepfake detection has expanded in scope with increasingly heterogeneous spoofing mechanisms, including speech synthesis, voice conversion, vocoder reconstruction, and neural-codec resynthesis. The resulting spoofing artifacts can be further shaped by variability in source speech, recording environments, and transmission channels. This variability makes robust generalization across heterogeneous conditions a central requirement for practical detection systems. This report presents Teffic-Audio, a general speech deepfake detection system designed for comprehensive evaluation environment. Teffic-Audio adopts a straightforward detector architecture consisting of a Conformer-based speech encoder, multi-head attentive statistics pooling, and a binary classifier. Rather than relying on additional architectural complexity, the system improves generalization through its training recipe, which integrates multi-source data, attack- and source-balanced sampling, and diverse audio augmentation. Trained only with open-source data, Teffic-Audio achieves a pooled EER of 1.454% on the 14 test sets of Speech-DF-Arena, outperforming all currently public systems on the leaderboard. It also obtains the lowest EER on five individual test sets and shows a favorable performance-complexity trade-off compared with larger leading systems. Overall, Teffic-Audio provides a strong and practical reference system for general speech deepfake detection.

6. Correlation between prosody and pragmatics: A case study of the discourse marker hālā `now’ in Persian

Authors: Soleiman Ghaderi, Moloud Asakereh, Kevin Tang

Categories: cs.CL | 35 pages, 0 figures Score: 5.40/10 (Obj:6 Id:5 Ind:8 Comp:6 Eff:5 Nov:4 Rep:6)

Strength: 方法严谨（混合效应模型+中介分析，90% 标注一致性，94 小时自定义 forced aligner），发现 textual 用法 duration 显著更短（β = -0.067, p = 0.001）、interactive 用法 intensity 显著更高（β = 1.665, p = 0.005），且 duration 效果经中介分析确认独立于 pitch/intensity。
Weakness: 新颖性为填补空白型——方法与 Kazemian et al. (2025) 对波斯语 bebin 的分析高度雷同，结论确认而非超越已有跨语言模式；未控制句法位置这一已知关键混淆变量；267 token 样本量小，pitch 模型可能统计功效不足。
Full review: Claude Code 全文七公理审稿

The Persian discourse marker hālā (‘now’) exhibits remarkable multifunctionality, extending far beyond its temporal adverbial role to encompass a variety of pragmatic functions. This study presents a pragmatic and acoustic analysis of hālā in spoken Persian, examining 267 instances from spontaneous conversations. While temporal uses were present, they were often combined with other discourse marker functions, indicating extensive multifunctionality, with 70% of tokens serving two or more pragmatic roles. Textual functions (topic shifting, signaling relationships, boundary marking, attention guidance, topic introduction, and topic emphasis) were most frequent, followed by interactive functions (turn management, listener engagement, and feedback regulation), and modal functions (epistemic stance, emotional expression, and attitudinal marking). Prosodic analysis revealed that duration and intensity are key cues for distinguishing hālā’s functions. Textual uses were significantly shorter, while temporal uses showed a tendency toward longer realizations. Interactive functions correlated with higher intensity, while modal functions showed a weaker tendency toward lower intensity. These findings indicate that duration and intensity are the main prosodic cues associated with functional differentiation in hālā, especially in textual and interactive uses.

7. Integrating Contextual Embeddings into Evaluation of Expressive MIDI Piano Performances

Authors: Dmitrii Gavrilev, Ilya Borovik, Vladimir Viro

Categories: cs.SD, cs.LG | Accepted at ISMIR 2026 Score: 5.30/10 (Obj:6 Id:5 Ind:8 Comp:5 Eff:5 Nov:4 Rep:8)

Strength: 首次在符号钢琴表演评测中系统对比 Aria/CLaMP3 嵌入并提出 KPD（per-score 条件化 MMD），velocity-transfer 实验证明 contextual 嵌入能捕捉属性级指标遗漏的分布偏移，Pereval 库开源且数据/模型全公开。
Weakness: 方法层为 KAD→符号域的换名迁移，per-sample 伪评分与人类 MOS 的 Kendall τB（0.44–0.51）未显著超越传统 IOI correlation，且未引用 arXiv:2601.19029 这一高度重叠的 6 个月前先前工作。
Full review: Claude Code 全文七公理审稿

Objective evaluation of expressive MIDI piano performances typically relies on attribute statistics such as timing, velocity, and duration of individual notes. However, these methods often disregard dependencies between notes, which poses a potential limitation in assessing the similarity between two sets of performances. In generative applications, the wide variety of expressive attributes makes it difficult to aggregate them into a single scalar metric for model selection. In this work, we reexamine attribute-scoped metrics and explore the perceptual properties of contextual embeddings from self-supervised symbolic music models, Aria and CLaMP3. Results from our listening study indicate that these models can be used as perceptual proxies, showing agreement with per-sample human ratings on par with traditional metrics. To measure conditional distributional similarity, we adapt Kernel Audio Distance to the symbolic music domain. Unlike Pearson correlation and reconstruction error, kernel-based methods on contextual embeddings do not require note alignment and are sensitive to contextual perturbations. To facilitate reproducibility, we release Pereval, an open-source library that integrates performance evaluation utilities, including both attribute-scoped and deep feature metrics.

Authors: Taewan Kim, Jiwoo Kang

Categories: cs.CV, cs.IR Score: 5.16/10 (Obj:8 Id:5 Ind:8 Comp:4 Eff:5 Nov:4 Rep:4)

Strength: 在 VoxCeleb 验证任务上实现 AUC 89.71%，相比 Chen 2023 (87.10%) 有持续提升，并在两个split、两个backbone和 AVSpeech 上保持一致。
Weakness: 核心方法是对其 SIGIR 2024 论文的简单线性插值扩展，且所有基线均截止至 2023 年，忽略了 2024-2025 年的竞争方法 (PAEFF, XM-ALIGN, FAME 挑战赛)。
Full review: Claude Code 全文七公理审稿

Face-and-voice association learning is one of the most challenging tasks in deep learning. In this paper, we propose a simple but powerful cross-modal feature embedding method for the association of faces and voices. Previous work has studied cross-modal association tasks to establish the correlation between voice clips and facial images. These works have addressed cross-modal discrimination but underestimate the importance of handling heterogeneity in inter-modal features between audio and video, resulting in a lot of false positives and false negatives. To tackle the problem, the proposed method learns the embeddings of cross-modal features by making another feature exist between cross-modal features, facilitating the voice and face features of the same person to be embedded in a convex hull. Moreover, the incorporation of cross-modal attention mechanisms with convex embedding techniques represents a highly effective strategy for the attenuation of false positives and false negatives, accomplished via the minimization of inter-class discrepancies. We exhaustively evaluated our method for cross-modal verification, matching, and retrieval tasks on the large-scale VoxCeleb dataset. Extensive experimental results demonstrate that the proposed method achieves notable improvements over existing state-of-the-art methods.

9. AgenticASR: Refining Speech Recognition in Real-World Scenarios via an Agentic Approach

Authors: Zixuan Jiang, Binghao Qiang, Jiaying Chi, Yanqiao Zhu, Kai Yu et al.

Categories: cs.AI | 15 pages, 3 figures, 14 tables Score: 4.47/10 (Obj:5 Id:5 Ind:5 Comp:2 Eff:5 Nov:2 Rep:5)

Strength: 提出 streaming 场景下 audio-to-clean-text 的 ASR–Refiner 架构并配双语 AASR-Bench，附 human–AI agreement 与多 ASR front-end 验证，方向实用。
Weakness: 与同组 2 个月内两篇前作高度重叠且未显式剥离增量，方法实为标准 streaming refinement 换名，缺最简 baseline 与公开强 baseline（DRES/Recording for Eyes）对比，benchmark judge 独立性证据不足。
Full review: Claude Code 全文七公理审稿

Automatic speech recognition (ASR) has achieved substantial gains in transcription accuracy, yet verbatim transcription does not necessarily produce readily usable text. It retains fillers, repetitions, false starts, and self-corrections that increase reading effort, obscure the speaker’s final intent, and propagate unresolved or abandoned content to downstream tasks. Existing spoken-to-written methods process completed audio or transcripts but cannot revise emitted text when later speech changes how preceding content should be interpreted. We therefore formulate Agentic Speech Recognition (AgenticSR), an audio-to-clean-text task that removes disfluencies, resolves self-corrections, and normalizes written form while preserving the speaker’s final intent. AgenticASR implements this task through an ASR–Refiner architecture that repeatedly transforms a bounded active context and replaces its corresponding output span as audio arrives. This enables continual emission and revision over streams of arbitrary duration. We also introduce AASR-Bench, a bilingual benchmark with fine-grained atomic rubrics. Across multiple ASR front ends, AgenticASR attains the highest AASR-Bench scores among evaluated systems. A human–AI agreement study shows that rubric-based judgments align with independent expert assessments. Ablations characterize Refiner capacity, context length, and the quality–latency trade-off between online and offline inference. Together, these results establish AgenticASR as a practical framework for intent-preserving clean transcription during ongoing speech. Code, AASR-Bench, and a demo will be released at https://github.com/AnXMuy/AgenticASR.