❌

Normal view

Received today — 7 October 2026 ⏭

Spatial, single-nucleus and pathological profiling of the invasive front in early hepatocellular carcinoma for characterizing specific leading-edge cell niche and improving recurrence modeling

✇Omics in Hepatocellular
By: Xicheng Wang · Xiaolan Mu · Yong Fu · Wei Dong · Wenfeng Lu · Xiuhua Li · Yanxiang Song · Changcheng Liu · Baoju Hou · Haibin Zhang · Xiong Chen · Yongchao Cai · Zhiying He
29 September 2026 at 18:00

Int J Biol Sci. 2026 Sep 10;22(14):8090-8118. doi: 10.7150/ijbs.137262. eCollection 2026.

ABSTRACT

The tumor leading edge (TLE) is a critical region where tumor cells interact with the microenvironment to drive invasion and metastasis; however, its cellular architecture in early hepatocellular carcinoma (HCC) remains poorly understood. Here, we integrated single-nucleus RNA-seq (snRNA-seq), spatial transcriptomics, and computational pathology to investigate TLE in early HCC. We annotated 35 cell subpopulations and identified STMN1-high tumor cells as a key malignant subset enriched at the invasive front, interacting with Treg, plasma B, LAMP3⁺ dendritic cells and SPP1⁺ macrophages. Spatial analysis revealed three co-localized cell pairs-(SPP1⁺ macrophages co-localized with Tip-like and inflammatory endothelial cells), (LAMP3⁺ DCs co-localized with naive T cells), and (plasma B cells co-localized with cancer-associated fibroblasts)-forming a leading-edge tumor microenvironment (L-TME) niche associated with early relapse. We developed an L-TME-related machine-learning benchmark framework incorporating 71 imaging features (65 deep-learning + 6 pathological) based on the snRNA-seq, spatial transcriptomics and pathomics. The pathology model achieved robust performance (mean C-index=0.77) and successfully predicted the recurrence of early HCC (log-rank p < 0.05) in TCGA (n=147) and an independent in-house cohort (n=123). This study delineates the TLE cellular ecosystem of early HCC, defines a spatially coordinated immunosuppressive L-TME niche, and provides a clinically applicable predictive tool for postoperative recurrence. Integrating multi-omics with computational pathology deepens our understanding of early HCC metastasis and offers insights into improved prognostication and therapeutic strategies.

PMID:42807944 | PMC:PMC13618224 | DOI:10.7150/ijbs.137262

Received — 10 September 2026 ⏭

Intranasal delivery of a vasoactive intestinal peptide-based circRNA vaccine induces systemic and mucosal immunity against RSV in mice

✇Molecular Therapy
By: Junqi Xiang · Jie Xu · Jing Zhu · Bo Yang · Mingling Wan · Jiangyu Cai · Shuying Chen · Yuncheng Wang · Shanrong Yang · Yaokai Wang · Li Lin · Qiqi Zhang · Na Zang · Enmei Liu · Yushun Wan
19 August 2026 at 08:00
A vasoactive intestinal peptide (VIP)-based protein carrier self-assembles with respiratory syncytial virus circular RNA vaccines for intranasal delivery, inducing systemic antibodies, mucosal IgA, and Th1-biased protection in mice. This platform offers a protein-guided strategy for respiratory mucosal RNA vaccination and broadens the application of VIP in vaccine delivery.

The redox architecture of gestational diabetes mellitus: from cellular stress engine to epigenetic and mitochondrial rewiring

✇(Multiomics OR Omics) AND (Pancreatic)
By: Haijuan Zhang · Shaoxiong Zhang · Yuanhao Liu · Jinjiang Shi · Aijun Yang · Ruirui Jin · Yuxia Jin · Xuya Wang · Yitao Fan · Qian Liu · Yanhua Zhang · Cheng Wang · Jianjun Mao · Zhiyuan Zheng · Ling Lv · Muzhou Teng
9 September 2026 at 18:00

Free Radic Biol Med. 2026 Sep 9;256:441-460. doi: 10.1016/j.freeradbiomed.2026.09.006. Online ahead of print.

ABSTRACT

Gestational diabetes mellitus (GDM) is a common pregnancy complication with a rising global prevalence, posing serious short-term and long-term health threats to both mothers and offspring. This review repositions GDM as a systemic disorder in which oxidative stress acts as a proposed mechanistic hub, linking upstream risk factors to downstream pathophysiology. We first examine how "upstream" factors-including genetic susceptibility, pre-conception status, and environmental exposures-converge to promote a state of pathological redox imbalance. We then examine key mechanistic pathways through which oxidative stress is thought to contribute to systemic insulin resistance and pancreatic β-cell failure, highlighting novel pathways involving intercellular communication via tunneling nanotubes and exosomes. Furthermore, we explore the downstream cascade, where oxidative stress may program maternal accelerated biological aging and multi-organ offspring disease trajectories through nuclear epigenetic programming and mitochondrial dysfunction programming, leaving what has been termed a persistent "metabolic memory". Consequently, this review evaluates emerging strategies that target oxidative stress for early prediction and precision intervention. Early prediction models based on direct redox biomarkers and multi-omics signatures hold potential to shift diagnosis from late-gestation oral glucose tolerance test (OGTT) to first-trimester risk stratification. Current supporting evidence draws from human epidemiological associations, ex vivo placental analyses, and experimental models. However, direct causal and interventional validation in pregnant women remains limited. Integrating targeted redox risk stratification and precision interventions into a life-course clinical framework may help interrupt the intergenerational transmission of metabolic disease initiated by GDM.

PMID:42716407 | DOI:10.1016/j.freeradbiomed.2026.09.006

Received — 27 May 2026 ⏭

STREAM: A Data-Centric Framework for Mining High-Value Task-Oriented Dialogues from Streaming Media

✇cs.AI, q-bio.NC updates on arXiv.org
By: Liang Xue · Haoyu Liu · Cheng Wang · Pengyu Chen · Haozhuo Zheng · Yang Liu
26 May 2026 at 12:00
arXiv:2605.25162v1 Announce Type: cross Abstract: Large language models for vertical domains are bottlenecked by the scarcity of complex, domain-specific task-oriented dialogues. Existing data acquisition pipelines face a persistent trilemma: expert annotation is expensive, real-world service conversations are constrained by privacy and commercial restrictions, and static corpora quickly become temporally stale. We propose Stream, a data-centric framework that leverages publicly available streaming media (live streams and short videos) to synthesize high-value service dialogues at scale. Stream mines authentic interaction signals from noisy streams and synthesizes conversations by integrating role-grounded persona construction with Conversational Blueprint construction; it further adopts retrieval-augmented generation (RAG) to support knowledge-aware responses. Based on Stream, we release StreamDial, a large-scale multi-domain dataset covering Automotive, Restaurant, and Hotel. StreamDial contains 87,498 dialogue sessions and 1,497,320 turns in total, with an average of 17.11 turns per session and a comparable scale across domains. Each session is organized as a structured quadruplet $\langle P_u, P_a, B, H \rangle$ that pairs dialogue history with explicit user/agent personas and a Conversational Blueprint, capturing realistic service behaviors such as requirement mining, constraint conflicts, negotiation, and recovery. Evaluations with automatic judges and downstream tasks show that StreamDial improves intrinsic dialogue quality over strong baselines, and models trained with StreamDial improve Dialogue State Tracking across backbones; we further report a completed human-evaluation set and encouraging multilingual transfer on Qwen3-8B under a controlled training budget. The data is released in https://github.com/hitxueliang/DialogDataSetBySTREAM.

Spatial transcriptomic-metabolic features of tumor foci and tumor capsule in microvascular invasion with hepatocellular carcinoma: A spatial multi-omics study

✇Omics in Hepatocellular
By: Zhi-Hui Luo · Na Wang · Jingwei Zhao · Fei Long · Si Wu · Wei Zhong · Wei-Ming Chen · Bicheng Wang · Kun Wang · Yufeng Yuan · Jingjiao Zhou · Chunhui Yuan · Fubing Wang
15 May 2026 at 18:00

PLoS Med. 2026 May 15;23(5):e1004703. doi: 10.1371/journal.pmed.1004703. eCollection 2026 May.

ABSTRACT

BACKGROUND: Microvascular invasion (MVI) is closely related to the recurrence and metastasis of hepatocellular carcinoma (HCC), but the underlying cellular mechanism remains largely elusive. This study aims to elucidate the regional cellular discrepancy between MVI-positive (MVI+) and MVI-negative (MVI-) HCC by integrating Spatial transcriptomics (ST) and spatial metabolomics (SM).

METHODS AND FINDINGS: ST and SM were performed on six tissue samples from four patients (including 2 MVI+, 2 MVI-, and 2 paratumor tissues), with the integration of 79 public single-cell RNA sequencing datasets of HCC. Patient identity was used as a covariate in the linear equation for regional differentially expressed gene analysis with the ST data. Clinical validation was conducted through multiplex immunofluorescence staining in 79 patients, together with external validation in the cancer genome atlas (TCGA)-liver hepatocellular carcinoma (LIHC) cohort (n = 299) and an independent microarray dataset (n = 62). For cell-type-specific metabolic profiling, spatial transcriptomic-metabolic registration was performed. The functional roles of key metabolites were further validated in vitro using inflammatory cancer-associated fibroblasts (iCAFs) derived from hepatic stellate cells (HSCs) and primary CAFs through co-culture models and various functional assays assessing cell proliferation, migration, and invasion. In the tumor lesion, a malignant STMN1+HMGN2+GPC3+ cell subtype enriched in MVI+ HCC was identified, which exhibited enhanced proliferative activity and was associated with poor prognosis. This finding was further confirmed in a local cohort of 79 patients, where multiplex immunofluorescence staining for the three genes (STMN1, HMGN2, and GPC3) showed significantly higher expression in the MVI+ group than in the MVI- group (p = 0.046). Integrated SM analysis further revealed that this cell population underwent metabolic reprogramming characterized by suppressed glycerolipid metabolism. In the tumor capsule, iCAFs-related genes were downregulated in MVI+ cases, and iCAFs were located distally from the tumor boundary. Spatial metabolite mapping showed a strong correlation between taurine and iCAFs, and functional assays demonstrated that taurine promotes HCC proliferation and migration by suppressing iCAF activity. One limitation of this study is the small sample size of spatial omics data, which hinders a more complete molecular functional analysis of the STMN1+HMGN2+GPC3+ cell subtype and iCAFs in MVI+ HCC. Larger-scale ST cohorts are required to further validate and expand the findings of this study.

CONCLUSIONS: This integrative spatial atlas proposes a hypothesis that there exists a highly proliferative and metabolically reprogrammed malignant cell subtype in the tumor lesion of MVI+ HCC, and that taurine in the tumor capsule modulates iCAF activity to influence tumor progression. The exploratory results provide mechanistic insights into MVI-related HCC progression and offer potential avenues for targeted therapeutic intervention of MVI+ HCC.

PMID:42139279 | PMC:PMC13178920 | DOI:10.1371/journal.pmed.1004703

Received — 13 April 2026 ⏭

Single-cell spatiotemporal dissection of the human maternal–fetal interface

✇Nature - Issue - nature.com science feeds
By: Cheng Wang · Yan Zhou · Yuejun Wang · Tuhin Kumar Guha · Zhida Luo · Anxhela Mustafaraj · Tara I. McIntyre · Marisa E. Schwab · Brittany R. Davidson · Gabriella C. Reeder · Ronald J. Wong · Sarah K. England · Juan M. Gonzalez · Robert Blelloch · Alexis J. Combes · Linda C. Giudice · Adrian Erlebacher · Tippi C. MacKenzie · David K. Stevenson · Gary M. Shaw · Michael P. Snyder · Xiaofei Sun · Virginia D. Winn · Susan J. Fisher · Jingjing Li
8 April 2026 at 08:00

Nature, Published online: 08 April 2026; doi:10.1038/s41586-026-10316-x

A single-cell multiomic atlas of the human maternal–fetal interface across pregnancy reveals cell types, states and spatial niches, developmental tissue architectures and transcriptional programmes, and identifies cell types with roles in pre-eclampsia, spontaneous preterm birth and miscarriage.
Received — 26 March 2026 ⏭

Beyond Matching to Tiles: Bridging Unaligned Aerial and Satellite Views for Vision-Only UAV Navigation

✇cs.AI, q-bio.NC updates on arXiv.org
By: Kejia Liu · Haoyang Zhou · Ruoyu Xu · Peicheng Wang · Mingli Song · Haofei Zhang
25 March 2026 at 12:00
arXiv:2603.22153v2 Announce Type: replace-cross Abstract: Recent advances in cross-view geo-localization (CVGL) methods have shown strong potential for supporting unmanned aerial vehicle (UAV) navigation in GNSS-denied environments. However, existing work predominantly focuses on matching UAV views to onboard map tiles, which introduces an inherent trade-off between accuracy and storage overhead, and overlooks the importance of the UAV's heading during navigation. Moreover, the substantial discrepancies and varying overlaps in cross-view scenarios have been insufficiently considered, limiting their generalization to real-world scenarios. In this paper, we present Bearing-UAV, a purely vision-driven cross-view navigation method that jointly predicts UAV absolute location and heading from neighboring features, enabling accurate, lightweight, and robust navigation in the wild. Our method leverages global and local structural features and explicitly encodes relative spatial relationships, making it robust to cross-view variations, misalignment, and feature-sparse conditions. We also present Bearing-UAV-90k, a multi-city benchmark for evaluating cross-view localization and navigation. Extensive experiments show encouraging results that Bearing-UAV yields lower localization error than previous matching/retrieval paradigm across diverse terrains. Our code and dataset will be made publicly available.
Received — 11 March 2026 ⏭

Mean Flow Policy with Instantaneous Velocity Constraint for One-step Action Generation

✇cs.AI, q-bio.NC updates on arXiv.org
By: Guojian Zhan · Letian Tao · Pengcheng Wang · Yixiao Wang · Yiheng Li · Yuxin Chen · Hongyang Li · Masayoshi Tomizuka · Shengbo Eben Li
10 March 2026 at 12:00
arXiv:2602.13810v2 Announce Type: replace-cross Abstract: Learning expressive and efficient policy functions is a promising direction in reinforcement learning (RL). While flow-based policies have recently proven effective in modeling complex action distributions with a fast deterministic sampling process, they still face a trade-off between expressiveness and computational burden, which is typically controlled by the number of flow steps. In this work, we propose mean velocity policy (MVP), a new generative policy function that models the mean velocity field to achieve the fastest one-step action generation. To ensure its high expressiveness, an instantaneous velocity constraint (IVC) is introduced on the mean velocity field during training. We theoretically prove that this design explicitly serves as a crucial boundary condition, thereby improving learning accuracy and enhancing policy expressiveness. Empirically, our MVP achieves state-of-the-art success rates across several challenging robotic manipulation tasks from Robomimic and OGBench. It also delivers substantial improvements in training and inference speed over existing flow-based policy baselines.
Received — 5 March 2026 ⏭

GreenPhase: A Green Learning Approach for Earthquake Phase Picking

✇cs.AI, q-bio.NC updates on arXiv.org
By: Yixing Wu · Shiou-Ya Wang · Dingyi Nie · Sanket Kumbhar · Yun-Tung Hsieh · Yun-Cheng Wang · Po-Chyi Su · C. -C. Jay Kuo
5 March 2026 at 13:00
arXiv:2603.03344v1 Announce Type: cross Abstract: Earthquake detection and seismic phase picking are fundamental yet challenging tasks in seismology due to low signal-to-noise ratios, waveform variability, and overlapping events. Recent deep-learning models achieve strong results but rely on large datasets and heavy backpropagation training, raising concerns over efficiency, interpretability, and sustainability. We propose GreenPhase, a multi-resolution, feed-forward, and mathematically interpretable model based on the Green Learning framework. GreenPhase comprises three resolution levels, each integrating unsupervised representation learning, supervised feature learning, and decision learning. Its feed-forward design eliminates backpropagation, enabling independent module optimization with stable training and clear interpretability. Predictions are refined from coarse to fine resolutions while computation is restricted to candidate regions. On the Stanford Earthquake Dataset (STEAD), GreenPhase achieves excellent performance with F1 scores of 1.0 for detection, 0.98 for P-wave picking, and 0.96 for S-wave picking. This is accomplished while reducing the computational cost (FLOPs) for inference by approximately 83% compared to state-of-the-art models. These results demonstrate that the proposed model provides an efficient, interpretable, and sustainable alternative for large-scale seismic monitoring.

Kaleido: Open-Sourced Multi-Subject Reference Video Generation Model

✇cs.AI, q-bio.NC updates on arXiv.org
By: Zhenxing Zhang · Jiayan Teng · Zhuoyi Yang · Tiankun Cao · Cheng Wang · Xiaotao Gu · Jie Tang · Dan Guo · Meng Wang
5 March 2026 at 13:00
arXiv:2510.18573v2 Announce Type: replace-cross Abstract: We present Kaleido, a subject-to-video~(S2V) generation framework, which aims to synthesize subject-consistent videos conditioned on multiple reference images of target subjects. Despite recent progress in S2V generation models, existing approaches remain inadequate at maintaining multi-subject consistency and at handling background disentanglement, often resulting in lower reference fidelity and semantic drift under multi-image conditioning. These shortcomings can be attributed to several factors. Primarily, the training dataset suffers from a lack of diversity and high-quality samples, as well as cross-paired data, i.e., paired samples whose components originate from different instances. In addition, the current mechanism for integrating multiple reference images is suboptimal, potentially resulting in the confusion of multiple subjects. To overcome these limitations, we propose a dedicated data construction pipeline, incorporating low-quality sample filtering and diverse data synthesis, to produce consistency-preserving training data. Moreover, we introduce Reference Rotary Positional Encoding (R-RoPE) to process reference images, enabling stable and precise multi-image integration. Extensive experiments across numerous benchmarks demonstrate that Kaleido significantly outperforms previous methods in consistency, fidelity, and generalization, marking an advance in S2V generation.
Received — 25 February 2026 ⏭

Decentralized Attention Fails Centralized Signals: Rethinking Transformers for Medical Time Series

✇cs.AI, q-bio.NC updates on arXiv.org
By: Guoqi Yu · Juncheng Wang · Chen Yang · Jing Qin · Angelica I. Aviles-Rivero · Shujun Wang
24 February 2026 at 13:00
arXiv:2602.18473v1 Announce Type: cross Abstract: Accurate analysis of medical time series (MedTS) data, such as electroencephalography (EEG) and electrocardiography (ECG), plays a pivotal role in healthcare applications, including the diagnosis of brain and heart diseases. MedTS data typically exhibit two critical patterns: temporal dependencies within individual channels and channel dependencies across multiple channels. While recent advances in deep learning have leveraged Transformer-based models to effectively capture temporal dependencies, they often struggle with modeling channel dependencies. This limitation stems from a structural mismatch: MedTS signals are inherently centralized, whereas the Transformer's attention mechanism is decentralized, making it less effective at capturing global synchronization and unified waveform patterns. To address this mismatch, we propose CoTAR (Core Token Aggregation-Redistribution), a centralized MLP-based module designed to replace decentralized attention. Instead of allowing all tokens to interact directly, as in standard attention, CoTAR introduces a global core token that serves as a proxy to facilitate inter-token interactions, thereby enforcing a centralized aggregation and redistribution strategy. This design not only better aligns with the centralized nature of MedTS signals but also reduces computational complexity from quadratic to linear. Experiments on five benchmarks validate the superiority of our method in both effectiveness and efficiency, achieving up to a 12.13% improvement on the APAVA dataset, while using only 33% of the memory and 20% of the inference time compared to the previous state of the art. Code and all training scripts are available at https://github.com/Levi-Ackman/TeCh.

ALOE: Action-Level Off-Policy Evaluation for Vision-Language-Action Model Post-Training

✇cs.AI, q-bio.NC updates on arXiv.org
By: Rushuai Yang · Hecheng Wang · Chiming Liu · Xiaohan Yan · Yunlong Wang · Xuan Du · Shuoyu Yue · Yongcheng Liu · Chuheng Zhang · Lizhe Qi · Yi Chen · Wei Shan · Maoqing Yao
24 February 2026 at 13:00
arXiv:2602.12691v2 Announce Type: replace-cross Abstract: We study how to improve large foundation vision-language-action (VLA) systems through online reinforcement learning (RL) in real-world settings. Central to this process is the value function, which provides learning signals to guide VLA learning from experience. In practice, the value function is estimated from trajectory fragments collected from different data sources, including historical policies and intermittent human interventions. Estimating the value function of current behavior quality from the mixture data is inherently an off-policy evaluation problem. However, prior work often adopts conservative on-policy estimation for stability, which avoids direct evaluation of the current high-capacity policy and limits learning effectiveness. In this paper, we propose ALOE, an action-level off-policy evaluation framework for VLA post-training. ALOE applies chunking-based temporal-difference bootstrapping to evaluate individual action sequences instead of predicting final task outcomes. This design improves effective credit assignment to critical action chunks under sparse rewards and supports stable policy improvement. We evaluate our method on three real-world manipulation tasks, including smartphone packing as a high-precision task, laundry folding as a long-horizon deformable-object task, and bimanual pick-and-place involving multi-object perception. Across all tasks, ALOE improves learning efficiency without compromising execution speed, showing that off-policy RL can be reintroduced in a reliable manner for real-world VLA post-training. Videos and additional materials are available at our project website.
Received — 17 February 2026 ⏭

Pailitao-VL: Unified Embedding and Reranker for Real-Time Multi-Modal Industrial Search

✇cs.AI, q-bio.NC updates on arXiv.org
By: Lei Chen · Chen Ju · Xu Chen · Zhicheng Wang · Yuheng Jiao · Hongfeng Zhan · Zhaoyang Li · Shihao Xu · Zhixiang Zhao · Tong Jia · Jinsong Lan · Xiaoyong Zhu · Bo Zheng
17 February 2026 at 13:00
arXiv:2602.13704v1 Announce Type: cross Abstract: In this work, we presented Pailitao-VL, a comprehensive multi-modal retrieval system engineered for high-precision, real-time industrial search. We here address three critical challenges in the current SOTA solution: insufficient retrieval granularity, vulnerability to environmental noise, and prohibitive efficiency-performance gap. Our primary contribution lies in two fundamental paradigm shifts. First, we transitioned the embedding paradigm from traditional contrastive learning to an absolute ID-recognition task. Through anchoring instances to a globally consistent latent space defined by billions of semantic prototypes, we successfully overcome the stochasticity and granularity bottlenecks inherent in existing embedding solutions. Second, we evolved the generative reranker from isolated pointwise evaluation to the compare-and-calibrate listwise policy. By synergizing chunk-based comparative reasoning with calibrated absolute relevance scoring, the system achieves nuanced discriminative resolution while circumventing the prohibitive latency typically associated with conventional reranking methods. Extensive offline benchmarks and online A/B tests on Alibaba e-commerce platform confirm that Pailitao-VL achieves state-of-the-art performance and delivers substantial business impact. This work demonstrates a robust and scalable path for deploying advanced MLLM-based retrieval architectures in demanding, large-scale production environments.

Mean Flow Policy with Instantaneous Velocity Constraint for One-step Action Generation

✇cs.AI, q-bio.NC updates on arXiv.org
By: Guojian Zhan · Letian Tao · Pengcheng Wang · Yixiao Wang · Yiheng Li · Yuxin Chen · Masayoshi Tomizuka · Shengbo Eben Li
17 February 2026 at 13:00
arXiv:2602.13810v1 Announce Type: cross Abstract: Learning expressive and efficient policy functions is a promising direction in reinforcement learning (RL). While flow-based policies have recently proven effective in modeling complex action distributions with a fast deterministic sampling process, they still face a trade-off between expressiveness and computational burden, which is typically controlled by the number of flow steps. In this work, we propose mean velocity policy (MVP), a new generative policy function that models the mean velocity field to achieve the fastest one-step action generation. To ensure its high expressiveness, an instantaneous velocity constraint (IVC) is introduced on the mean velocity field during training. We theoretically prove that this design explicitly serves as a crucial boundary condition, thereby improving learning accuracy and enhancing policy expressiveness. Empirically, our MVP achieves state-of-the-art success rates across several challenging robotic manipulation tasks from Robomimic and OGBench. It also delivers substantial improvements in training and inference speed over existing flow-based policy baselines.

GSRM: Generative Speech Reward Model for Speech RLHF

✇cs.AI, q-bio.NC updates on arXiv.org
By: Maohao Shen · Tejas Jayashankar · Osama Hanna · Naoyuki Kanda · Yancheng Wang · Kate\v{r}ina \v{Z}mol\'ikov\'a · Ruiming Xie · Niko Moritz · Anfeng Xu · Yashesh Gaur · Gregory Wornell · Qing He · Jilong Wu
17 February 2026 at 13:00
arXiv:2602.13891v1 Announce Type: cross Abstract: Recent advances in speech language models, such as GPT-4o Voice Mode and Gemini Live, have demonstrated promising speech generation capabilities. Nevertheless, the aesthetic naturalness of the synthesized audio still lags behind that of human speech. Enhancing generation quality requires a reliable evaluator of speech naturalness. However, existing naturalness evaluators typically regress raw audio to scalar scores, offering limited interpretability of the evaluation and moreover fail to generalize to speech across different taxonomies. Inspired by recent advances in generative reward modeling, we propose the Generative Speech Reward Model (GSRM), a reasoning-centric reward model tailored for speech. The GSRM is trained to decompose speech naturalness evaluation into an interpretable acoustic feature extraction stage followed by feature-grounded chain-of-thought reasoning, enabling explainable judgments. To achieve this, we curated a large-scale human feedback dataset comprising 31k expert ratings and an out-of-domain benchmark of real-world user-assistant speech interactions. Experiments show that GSRM substantially outperforms existing speech naturalness predictors, achieving model-human correlation of naturalness score prediction that approaches human inter-rater consistency. We further show how GSRM can improve the naturalness of speech LLM generations by serving as an effective verifier for online RLHF.
There are no more articles
❌