论文检索

Siddhant Arora, Jinchuan Tian, Jiatong Shi, Hayato Futami, Yosuke Kashiwagi, Emiru Tsunoo, Shinji Watanabe

Reinforcement learning from human or AI feedback (RLHF/RLAIF) for speech-in/speech-out dialogue systems (SDS) remains underexplored, with prior work largely limited to single semantic rewards applied at the utterance level. Such setups overlook the multi-dimensional and multi-modal nature of conversational quality, which encompasses semantic coherence, audio naturalness, speaker consistency, emotion alignment, and turn-taking behavior. Moreover, they are fundamentally mismatched with duplex spoken dialogue systems that generate responses incrementally, where agents must make decisions based on partial utterances. We address these limitations with the first multi-reward RLAIF framework for SDS, combining semantic, audio-quality, and emotion-consistency rewards. To align utterance-level preferences with incremental, blockwise decoding in duplex models, we apply turn-level preference sampling and aggregate per-block log-probabilities within a single DPO objective. We present the first systematic study of preference learning for improving SDS quality in both multi-turn Chain-of-Thought and blockwise duplex models, and release a multi-reward DPO dataset to support reproducible research. Experiments show that single-reward RLAIF selectively improves its targeted metric, while joint multi-reward training yields consistent gains across semantic quality and audio naturalness. These results highlight the importance of holistic, multi-reward alignment for practical conversational SDS.

Han Zhu, Wei Kang, Liyong Guo, Zengwei Yao, Fangjun Kuang, Weiji Zhuang, Zhaoqing Li, Zhifeng Han, Dong Zhang, Xin Zhang 等

Generating spoken dialogue is inherently more complex than monologue text-to-speech (TTS), as it demands both realistic turn-taking and the maintenance of distinct speaker timbres. While existing autoregressive (AR) models have made progress, they often suffer from high inference latency and stability issues. To overcome these limitations, we propose ZipVoice-Dialog, a non-autoregressive (NAR) zero-shot spoken dialogue generation model based on flow-matching. Observing that applying vanilla flow-matching to dialogue generation leads to poor speech intelligibility and turn-taking precision, we introduce two simple yet effective methods to adapt flow-matching architectures for dialogue generation: (1) a curriculum learning strategy to ensure robust speech-text alignment, and (2) speaker-turn embeddings to govern precise speaker turn-taking. Additionally, we introduce dedicated strategies to support stereo dialogue generation.Recognizing the lack of training datasets in this field, we curate and release OpenDialog, the first large-scale (6.8k hours) open-source spoken dialogue dataset derived from in-the-wild speech data. Moreover, for fair and rigorous evaluations, we established a benchmark to comprehensively evaluate dialogue generation models. Experiments demonstrate the effectiveness of the proposed methods and dataset, showing that ZipVoice-Dialog achieves superior performance in inference speed, intelligibility, speaker turn-taking accuracy, and speaker similarity. Our code, model checkpoints, and the OpenDialog dataset are publicly available.

Chan-Jan Hsu, Liang-Hsuan Tseng, Yi-Cheng Lin, Yen-Chun Kuo, Ju-Chieh Chou, Kai-Wei Chang, Hung-yi Lee, Carlos Busso

Generative spoken language models pretrained on large-scale raw audio can continue a speech prompt with appropriate content while preserving attributes like speaker and emotion, serving as foundation models for spoken dialogue. In prior literature, these models are often evaluated using “global token perplexity”, which directly applies the text perplexity formulation to speech tokens. However, this practice overlooks fundamental differences between speech and text modalities, possibly leading to an underestimation of the speech characteristics. In this work, we propose a variety of likelihood- and generative-based evaluation methods that serve in place of naive global token perplexity. We demonstrate that the proposed evaluations more faithfully reflect perceived generation quality, as evidenced by stronger correlations with human-rated mean opinion scores (MOS). When assessed under the new metrics, the relative performance landscape of spoken language models is reshaped, revealing a significantly reduced gap between the best-performing model and the human topline. Together, these results suggest that appropriate evaluation is critical for accurately assessing progress in spoken language modeling.

Xuan Luo, Lewei Yao, Lanqing Hong, Kai Chen, Dehua Tao, Daxin Tan, Yukun Deng, Ruifeng Xu, Jing Li

While the automatic evaluation of omni-modal large models (OLMs) is essential, assessing empathy remains a significant challenge due to its inherent affectivity. To investigate this challenge, we introduce AEQ-Bench (Audio Empathy Quotient Benchmark), a novel benchmark to systematically assess two core empathetic capabilities of OLMs: (i) generating empathetic responses by comprehending affective cues from multi-modal inputs (audio + text), and (ii) judging the empathy of audio responses without relying on text transcription. Compared to existing benchmarks, AEQ-Bench incorporates two novel settings that vary in context specificity and speech tone. Comprehensive assessment across linguistic and paralinguistic metrics reveals that (1) OLMs trained with audio output capabilities generally outperformed models with text-only outputs, and (2) while OLMs align with human judgments for coarse-grained quality assessment, they remain unreliable for evaluating fine-grained paralinguistic expressiveness.

Da Shen, Zhenqiang Weng, Tianyu Liu, Gongyu Chen, Runhua Shi, Jiahui Chen, Chaofan Ding, Wei-Qiang Zhang, Zihao Chen

Modern speech and multimodal generation systems, such as singing voice conversion and audio-driven lip synchronization, critically depend on temporally stable and semantically unambiguous vocal representations. In practical pipelines, such representations are typically derived from music source separation (MSS) applied to mixed musical recordings. However, standard MSS paradigms often aggregate lead vocals and backing harmonies into a single vocal stream. Although multi-stem separation has been explored, existing approaches remain primarily optimized for signal-level reconstruction, often overlooking the intricate structural disentanglement required by downstream generation tasks. From a generation-oriented perspective, this motivates revisiting vocal separation from a representation learning standpoint. To this end, we propose VocalRep, a structure-aware learning framework designed to disentangle lead vocals, harmonies, and accompaniment while enforcing role consistency across long-form audio. By integrating global vocal identity conditioning with ranking-based objectives, VocalRep extracts role-consistent lead vocal representations without relying on explicit pitch or symbolic annotations. Experimental results demonstrate that VocalRep significantly improves performance in downstream singing voice conversion and audio-driven lip synchronization.

Soham Dan, Himanshu Beniwal, Thomas Hartvigsen

Large language models (LLMs) are transforming natural language processing across diverse linguistic communities. However, they can reproduce and amplify toxic content, including hate speech, harassment, and bias, posing significant risks to multilingual applications. We provide the first comprehensive survey of the many detoxification methods specifically tailored to multilingual LLMs. First, we define toxicity its measurement, then we provide a brief review of monolingual mitigation strategies, including data filtering, style transfer, expert-based logit steering, retrieval augmentation, and alignment with human feedback. We then present an in-depth taxonomy of multilingual approaches spanning (1) training methods, (2) post-hoc editing and decoding strategies, (3) alignment and reinforcement-learning techniques, and (4) data-centric innovations, such as parallel detox corpora and synthetic data generation. Finally, we discuss open challenges in multilingual detoxification, including data scarcity, evaluation inconsistencies, cultural nuances and biases. Overall, we produce a needed overview of the state of multi-lingual toxicity detection and mitigation on which the community can ground to build globally safe and equitable LLMs.

Jieyi Wang, Yazhe Niu, Dexuan Xu, Zhongyu Wei

Recent Large Audio Language Models (LALMs) have shown strong capabilities in audio understanding, yet their reasoning remains vulnerable to perceptual errors, especially in noisy and multi-speaker environments. We argue that reliable audio reasoning requires first grounding model’s perception in structured auditory scenes. Motivated by Auditory Scene Analysis, we introduce **PAQA**, a large-scale dataset for **Perception-Aware Question Answering** covering over 300 categories. PAQA adopts a hierarchical decoupling strategy that separates speech from environmental sounds and distinguishes among multiple speakers, providing explicit perceptual supervision for audio reasoning. Building on this, we propose **HyPeR**, a two-stage **Hybrid Perception-Reasoning** framework for perception-grounded audio understanding. In Stage I, the model is fine-tuned on PAQA for cold start to improve perception of acoustic attributes in complex auditory scenes. In Stage II, we further refine its internal reasoning via **Group Relative Policy Optimization (GRPO)**. To support deliberation under acoustic ambiguity, we introduce **PAUSE tokens** for latent computation and a **Perceptual Consistency Reward** to align reasoning rationales with the underlying audio evidence. Extensive ablation studies isolate the effects of the perception-attention mechanism, self-correction module, and pause-based reasoning strategy. Experiments on multiple benchmarks show that HyPeR consistently improves over the base model, including on MMAU-mini (+13.1%), MMAR (+25.5%), and PAQA (+28.2%), while achieving performance comparable to much larger models. Additional analyses of inference latency and computational overhead show that these gains come with acceptable efficiency trade-offs. Overall, our results demonstrate the effectiveness of hybrid perception-grounded reasoning for robust audio understanding.

Michael S. Yantosca, Albert M. K. Cheng

Recent automated transcription systems have focused on end-to-end orthographic approaches driven by deep neural networks and sequence-to-sequence transformers. Growing public interest in transcription at the phonemic or phonetic level has led to re-purposing these systems to segment and identify phones, the basic sounds which comprise human speech. However, they miss the mark on a fundamental component of time-series analysis, namely time. For linguistic applications which require high fidelity in the temporal domain, the loss of timing information is untenable. Our work proposes a deadline-bounded expectation maximization (EM) algorithm with a novel initialization method to estimate formants, i.e., salient speech frequencies, for enhanced phonetic segmentation. Based on the concept of spectral gravity, i.e., treating spectral energy as mass attenuated by the square of frequency distance across the spectrum, our technique outperforms the recent state of the art on key clustering metrics, generating reasonable alignments across multiple languages with no a priori training.

Mohd Mujtaba Akhtar, Girish, Muskaan Singh

In this study, we present Healthcare Codec-Fake Detection (HCFD), a new task for detecting codec-fakes under pathological speech conditions. We intentionally focus on codec based synthetic speech in this work, since neural codec decoding forms a core building block in modern speech generation pipelines. First, we release Healthcare CodecFake, the first pathology-aware dataset containing paired real and NAC-synthesized speech across multiple clinical conditions and codec families. Our evaluations show that SOTA codec-fake detectors trained primarily on healthy speech perform poorly on Healthcare CodecFake, highlighting the need for HCFD-specific models. Second, we demonstrate that PaSST outperforms existing speech-based models for HCFD, benefiting from its patch-based spectro-temporal representation. Finally, we propose PHOENIX-Mamba, a geometry-aware framework that models codec-fakes as multiple self-discovered modes in hyperbolic space and achieves the strongest performance on HCFD across clinical conditions and codecs. Experiments on HCFK show that PHOENIX-Mamba (PaSST) achieves the best overall performance, reaching 97.04 Acc on E-Dep, 96.73 on E-Alz, and 96.57 on E-Dys, while maintaining strong results on Chinese with 94.41 (Dep), 94.40 (Alz), and 93.20 (Dys). This geometry-aware formulation enables self-discovered clustering of heterogeneous codec-fake modes in hyperbolic space, facilitating robust discrimination under pathological speech variability. PHOENIX-Mamba achieves topmost performance on the HCFD task across clinical conditions and codecs.

Xi Chen, Wei Xue, Yike Guo

Role-playing has garnered rising attention as it provides a strong foundation for human-machine interaction and facilitates sociological research. However, current work is confined to textual modalities, neglecting speech, which plays a predominant role in daily life, thus limiting genuine role-playing. To bridge this gap, we conceptualize and benchmark speech role-playing through ActorMindBench, and we present a corresponding reasoning framework, called ActorMind. Specifically, (1) Speech Role-Playing enables models to deliver spontaneous responses with personalized verbal traits based on their role, the scene, and spoken dialogue. (2) ActorMindBench is a hierarchical benchmark comprises Utterance-Level content with 7,653 utterances, Scene-Level content with 313 scenes, and Role-Level content with 6 roles. (3) ActorMind is an off-the-shelf, multi-agent, chain-of-though style reasoning framework that emulates how human actors perform in theaters. Concretely, ActorMind first reads its assigned role description via Eye Agent, then comprehends emotional cues within contextual spoken dialogues through Ear Agent. Subsequently, Brain Agent generates a descriptive emotional state, and finally, Mouth Agent delivers the scripts infused with corresponding emotion state. Experimental results demonstrate the effectiveness of ActorMind in enhancing speech role-playing. The project page is available at https://github.com/OzymandiasChen/ActorMind.

Yaya SY, Dioula Doucouré, Christophe Cerisara, Irina Illina

Speakers of under-represented languages face both a language barrier, as most online knowledge is in a few dominant languages, and a modality barrier, since information is largely text-based while many languages are primarily oral. We address this for French-Wolof by training the first bilingual speech-text Matryoshka embedding model, enabling efficient retrieval of French text from Wolof speech queries without relying on a costly ASR-translation pipelines. We introduce large-scale data curation pipelines and new benchmarks, compare modeling strategies, and show that modality fusion within a frozen text Matryoshka model performs best. Although trained only for retrieval, the model generalizes well to other tasks, such as speech intent detection, indicating the learning of general semantic representations. Finally, we analyze cost-accuracy trade-offs across Matryoshka dimensions and ranks, showing that information is concentrated only in a few components, suggesting potential for efficiency improvements.

Francielle Vargas, Jackson Trager, Diego Alves, Matteo Guida, Surendrabikram Thapa, Berk Atıl, Daryna Dementieva, Andrew J Smart, Ameeta Agrawal

Existing hate speech detection models are often opaque and rely on surface-level lexical cues, which makes them vulnerable to spurious correlations and limits robustness, interpretability and cultural contextualization. We propose Supervised Moral Rationale Attention (SMRA), the first self-explaining hate speech detection framework to incorporate moral rationales as direct supervision for attention alignment. Based on Moral Foundations Theory, SMRA aligns token-level attention with expert-annotated moral rationales, guiding models to attend to morally salient spans. Unlike prior rationale-supervised or post-hoc approaches, SMRA integrates moral rationale supervision directly into the training objective, producing inherently interpretable and contextualized explanations. To support our framework, we also introduce HateBRMoralXplain, a Brazilian Portuguese benchmark dataset annotated with hate labels, moral categories, token-level moral rationales, and socio-political metadata. Across binary hate speech detection and multi-label moral sentiment classification, SMRA consistently improves performance while enhancing both faithful and plausible explanations. Although explanations become more concise, sufficiency decreases, indicating more compact and informative rationales. Fairness remains stable, suggesting that improvements in explanation quality do not introduce significant bias trade-offs.

Yanshan Liu, Hongbo Zhang, Zhen Sun, Jiaheng Wei, Kaishun Wu

In complex domains like interior design, user requests are often ambiguous and multimodal. Professional designers address this by asking strategic clarification questions based on hierarchical priorities, a capability lacking in current Vision-Language Models (VLMs). When fine-tuned on dialogue data, existing models often exhibit modality forgetting, overfitting to textual patterns while neglecting visual cues and thus producing hallucinated or visually irrelevant questions. To bridge this gap, we introduce VIDA (Visual Intent-driven Design Assistant), an assistant designed to generate proactive, visually grounded, and strategically prioritized clarification questions. Instead of standard fine-tuning, we propose a strategy-aware alignment framework that evolves from imitation learning to value-driven reinforcement. We utilize Group Sequence Policy Optimization to strictly enforce expert protocols, ensuring the model not only mimics fluent speech but also adheres to optimal inquiry strategies. Crucially, we design a novel hierarchical reward mechanism with Dynamic Intent Binding to align the assistant with professional prioritization standards. To facilitate this research, we construct and release InteriorClarify, a multimodal benchmark dataset comprising 1,016 real-world consultation cases annotated with this three-tier intent hierarchy. Extensive experiments demonstrate that VIDA sets a new state-of-the-art, improving the Strategic Alignment Score (SAS) by 20.59% over SFT baselines and effectively restoring visual grounding capabilities lost during standard fine-tuning.

Felix Herron, Solange Rossato, Alexandre Allauzen, François Portet

Modern automatic speech recognition (ASR) systems have been observed to function better for certain speaker groups (SGs) than others, despite recent gains in overall performance. One potential impediment to progress towards fairer ASR is a more nuanced understanding of the types of modeling errors that speech encoder models make, and in particular the difference between the structure of embeddings for high-performance and low-performance SGs. This paper proposes a framework typifying two types of error that can occur in modeling phonemes in ASR systems: random error/high variance in phoneme embedding, vs systematic error/embedding bias. We find that training phoneme classification probes only on a single, typically disadvantaged SG, sometimes improves performance for that SG, which is evidence for the existence of SG-level bias in phoneme embeddings. On the other hand, we find that speakers and SGs with higher levels of phoneme variance are the same as those with worse phoneme prediction accuracy. We conclude that both types of error are present in phoneme embeddings and both are candidate causes for SG-level unfairness in ASR, though random error is likely a greater hindrance to fairness than systematic error. Furthermore, we find that finetuning encoder models using a fairness-enhancing algorithm (domain enhancing and adversarial training) changes neither the benefits of in-domain phoneme classification probe training, nor measured levels of random embedding error.

Sanne Hoeken, Sophie Jasmin Spliethoff, Silke Schwandt, Özge Alacam, Sina Zarrieß

Research on hate speech detection (HSD) has centered on modern data, even though offensive language has a much longer history. This paper presents the first systematic evaluation of instruction-tuned LLMs on Early Modern English invectives, compared with a modern hate-speech benchmark. Our work applies a modular prompt design to measure the contribution of definitional richness, contextual grounding, decision rules and few-shot examples. The results indicate that clearer annotation boundaries in the curated historical corpus lead to higher classification performance compared to the modern benchmark, despite the disadvantage of linguistic unfamiliarity. Prompt brittleness, however, persists across both domains. Classification-oriented components (rules, examples) drive the strongest effects, while definitional or contextual additions matter less. Fine-tuned encoder models still outperform LLMs, but some prompt configurations can narrow the gap. Overall, our study provides practical guidance for prompt design in both digital humanities and HSD and new opportunities for tracing the historical development of hate speech.

Jiliang Hu, Wenfu Wang, Zuchao Li, Chenxing Li, Yiyang Zhao, Hanzhao Li, Liqiang Zhang, Meng Yu, Dong Yu

While large audio language models (LALMs) have driven significant progress in multimodal conversational systems, current benchmarks suffer from critical limitations: they are largely English-centric, use synthetic speech, and fail to provide comprehensive, discriminative evaluation across key dimensions. To fill this gap, we present Voice Chat Bot Bench (VCB Bench), a novel, high-quality Chinese benchmark built exclusively on real human speech. VCB Bench assesses LALMs across three complementary axes: instruction following (including speech-level control beyond text commands), knowledge understanding (including general knowledge, reasoning, and daily dialogue), and robustness (evaluating stability under variations in content, environment, and speaker characteristics). Experiments conducted on representative LALMs reveal notable performance disparities and offer tangible insights for future improvements. VCB Bench serves as a reproducible and fine-grained framework, providing standardized evaluation and practical guidance for the development of Chinese voice conversational models.

Yifei Cao, Changhao Jiang, Jiabao Zhuang, Jiajun Sun, Ming Zhang, Zhiheng Xi, Hui Li, Shihan Dou, Yuran Wang, Yunke Zhang 等

Speech quality assessment (SQA) is typically formulated as a score regression task based on subjective ratings, such as the Mean Opinion Score (MOS), which inherently suffer from inconsistent standards and limit cross-dataset training and evaluation. To address these limitations, we reformulate SQA as a preference-based comparison paradigm and construct MOS-Pref, a large-scale MOS-derived preference dataset. Building on MOS-Pref, we systematically implement and evaluate three reward modeling paradigms: scalar, semi-scalar, and generative reward models, alongside existing SQA approaches. Our experiments reveal three key findings: (1) scalar models achieve the strongest overall performance, consistently exceeding 74% accuracy; (2) score regression-based approaches generally underperform preference-based methods in both overall performance and generalization; and (3) all reward models struggle on pairs with very small MOS gap. Motivated by these observations, we propose a MOS-aware GRM design that incorporates MOS gap into the reward function during reinforcement learning. Experimental results show that the MOS-aware GRM significantly improves fine-grained speech quality discrimination. We hope this work fosters more rigorous and scalable research in SQA.

Kwok-Ho Ng, Tingting Song, Yongdong WU, Zhihua Xia

Advanced speech synthesis technologies have enabled highly realistic speech generation, posing security risks that motivate research into audio deepfake detection (ADD). While state space models (SSMs) offer linear complexity, pure causal SSMs architectures often struggle with the content-based retrieval required to capture global frequency-domain artifacts. To address this, we explore the scaling properties of hybrid architectures by proposing XLSR-MamBo, a modular framework integrating an XLSR front-end with synergistic Mamba-Attention backbones. We systematically evaluate four topological designs using advanced SSM variants, Mamba, Mamba2, Hydra, and Gated DeltaNet. Experimental results demonstrate that the MamBo-3-Hydra-N3 configuration achieves competitive performance compared to other state-of-the-art systems on the ASVspoof 2021 LA, DF, and In-the-Wild benchmarks. This performance benefits from Hydra’s native bidirectional modeling, which captures holistic temporal dependencies more efficiently than the heuristic dual-branch strategies employed in prior works. Furthermore, evaluations on the DFADD dataset demonstrate robust generalization to unseen diffusion- and flow-matching-based synthesis methods. Crucially, our analysis reveals that scaling backbone depth effectively mitigates the performance variance and instability observed in shallower models. These results demonstrate the hybrid framework’s ability to capture artifacts in spoofed speech signals, providing an effective method for ADD. Codes are publicly available at https://github.com/saki-ciallo/XLSR-MamBo.

Zhenhua Xu, Dongsheng Chen, Shuo Wang, Jian Li, Chengjie Wang, Meng Han, Yabiao Wang

LLM role-playing seeks to portray arbitrary characters in interactive narratives, yet existing systems often lack immersion and adapt ability. They typically under-model dynamic environment information and assume a largely static scene/cast, offering limited support for multi-character orchestration, scene transitions, and on-the-fly character introduction. We propose an adaptive multi-agent interaction framework dubbed AdaMARP, which featuring an immersive message format that interleaves [Thought], (Action), Environment, and Speech, and an explicit Scene Manager that controls role-playing via discrete actions (init_scene, pick_speaker, switch_scene, add_role, end) with rationales. To train these abilities, we construct AdaRPSet for the Actor Model and AdaSMSet for supervising or chestration decisions, and introduce AdaptiveBench for trajectory-level evaluation. Experiments across multiple backbones and scales show consistent gains: AdaRPSet improves character consistency, environment grounding, and narrative coherence—an 8B actor outperforming several commercial LLMs, while AdaSMSet enables smoother scene transitions and more natural role introductions, surpassing Claude Sonnet 4.5 with only 14B LLMs.

Jaehee Kim, Ji Hoon Chung, Seoyoon Park, Unsol Kim, Kyungwon Park, JiHak Kim, Yi-Jun Chen, Hansaem Kim

Indirect speech acts (ISAs) require pragmatic reasoning over context, as directive intent cannot be inferred from surface form alone. Prior text-based studies and existing multimodal benchmarks largely overlook this requirement, focusing instead on explicitly encoded context or perceptual recognition, and thus underexplore context-dependent pragmatic understanding—particularly in high-context languages such as Korean. We introduce READI, a multimodal benchmark for evaluating ISA understanding through integrated reasoning over visual context and dialogue. READI models graded indirectness grounded in pragmatic theory and formulates the task as vision-based pragmatic question answering (V-PQA), supporting cross-lingual evaluation in English and Korean. Experiments show that even state-of-the-art multimodal models struggle with visually grounded indirect speech acts, with performance declining as indirectness increases, underscoring the need for benchmarks that explicitly target contextual pragmatic reasoning.