arXiv_SD
-
Preserving Speech-to-Text LLM Capabilities in Speech-to-Speech Generation
2026-06-29 21:55:18Yuxuan Hu, Heng Lu, Ruchao Fan, Yao Qian, Xiaofei Wang, Jian Xue, Heming Wang, Shuohang Wang, Young Jin Kim, Yelong Shen, Jinyu LiAbstract
Strong speech-to-text (S2T) LLMs already provide robust speech perception and text reasoning, but adding speech-to-speech (S2S) output is challenging: fine-tuning the backbone can degrade the original S2T performance, while attaching a downstream talker reintroduces a serial text-to-speech bottleneck. We present PRIME-Speech, a frozen-backbone S2S conversion framework that trains only speech-generation modules. PRIME-Speech synchronizes a causal audio post-decoder with intermediate hidden states of the frozen backbone, so codec tokens are generated from the model's evolving reasoning trajectory rather than from completed text chunks. The post-decoder uses mixed hidden-state, text, and audio-history conditioning, and a training-time packing strategy with turn-level audio KV-cache and position reset stabilizes multi-turn spoken interaction without additional multi-turn S2S training data. Multi-token prediction further reduces the effective codec prediction rate and improves first-audio latency without modifying the reasoning path. Across speech translation, spoken QA, speech understanding, and multi-turn dialogue, PRIME-Speech preserves the S2T behavior of the frozen backbone while producing accurate, low-WER spoken responses.
Abstract (translated)
URL
https://arxiv.org/abs/2606.30944
PDF
-
SyncCache: Exploiting Asymmetric Dynamics for Fast Audio-Driven Portrait Animation
2026-06-29 19:26:13Juncheng Ma, Yuxuan Du, Yanan Sun, Zhening Xing, Changlin Li, Zhenyu Tang, Bo Li, Peng-Tao Jiang, Li Yuan, Daquan Zhou, Yonghong TianAbstract
Diffusion Transformers (DiTs) have significantly advanced audio-driven portrait animation, but their high computational cost leads to substantial inference latency. Although training-free diffusion caching accelerates inference significant, existing methods are primarily developed for text-conditioned generation and overlook the spatial and modality imbalances inherent in audio-driven portrait animation. In this paper, we propose SyncCache, a training-free caching acceleration method tailored for DiT-based portrait animation that explicitly exploits asymmetric dynamics. Specifically, high-frequency dynamics driven by audio conditions and concentrated in human regions are more challenging and critical to cache and reuse than the low-frequency visual background in portrait animation. First, we introduce Spatially-Asymmetric Probing to prioritize error sensitivity in dynamic human region. Second, through Modality-Decoupled Caching, we bypass heavy DiT block by reusing stable inter-block residuals, while continuously recomputing lightweight audio blocks to preserve precise lip synchronization. Furthermore, we introduce a cache ratio to control cache capacity and formulate memory-adaptive cache selection as an offline dynamic programming problem without online overhead. Extensive experiments demonstrate that SyncCache achieves superior speed-quality trade-offs, delivering up to 4.12x acceleration on HunyuanVideo-Avatar and 3.75x on Wan-S2V with near-lossless visual fidelity and precise audio alignment.
Abstract (translated)
URL
https://arxiv.org/abs/2606.30849
PDF
-
AVTok: 1D Unified Tokenization for Holistic Audio-Video Generation
2026-06-29 18:35:17Kien T. Pham, I Chieh Chen, Qifeng Chen, Long ChenAbstract
Audio-video generation has recently gained unprecedented research attention, aiming to synthesize high-quality sounding video content with fine-grained synchronization and semantic alignment between the auditory and visual components. The preceding methods predominantly adopt a dual-branch design with separate tokenization and generation modules per modality, neglecting the representation gap while necessitating intensive computational resources for proper training. Inspired by recent advancements in one-dimensional visual tokenization, we present \textbf{AVTok}, a novel unified tokenizer designated for holistic audio-video generation. AVTok features a dual-stream transformer-based architecture with shared encoder-decoder and modal-specific learnable queries to efficiently and effectively encode an audio-video pair into a compact one-dimensional latent representation with a unified codebook. To cope with the heterogeneous information imbalance that hinders AVTok from exploiting aligned audio-visual information, we devise a hierarchical training strategy to progressively realize reconstruction capabilities for each modality. Extensive experiments demonstrate that AVTok excels both in audio-video reconstruction and when integrated into downstream pipelines for audio-to-video, video-to-audio, and class-conditional joint audio-video generation. AVTok paves the way for the challenge of joint audio-video tokenization and provides a potential direction to build unified large multimodal models for audio-video generation.
Abstract (translated)
URL
https://arxiv.org/abs/2606.30811
PDF
-
Probing-Guided Layer Selection from Self-Supervised Speech Models for Generalizable Audio Deepfake Detection
2026-06-29 18:19:37Marjan Beheshti, Majid Rostami, Bo ChenAbstract
Audio deepfake detection systems often fail to generalize across domains because they rely on features tied to specific attacks or recording conditions. Self-supervised speech models offer rich multi-layer representations, yet existing approaches either use a single layer or fuse all layers indiscriminately, and only reveal layer importance after training. We propose a model-agnostic, two-stage methodology that identifies informative depth zones before any task-specific model is trained. In the first stage, lightweight XGBoost probes evaluate each transformer layer's cross-domain discriminative power, producing a layer ranking. In the second stage, a compact neural classifier fuses only the selected layers through per-layer attention pooling and a shared bottleneck projection, while the backbone remains frozen. Applied across three backbones, the probing reveals two key findings. First, informative layers cluster in depth zones rather than at uniquely optimal positions: within-zone substitutions fall within multi-seed noise, while zone violations degrade performance by up to 5x. Second, the probing produces backbone-specific selections rather than a fixed layer recipe. On XLS-R-300M, four probing-selected layers with 1.34M trainable parameters achieve 4.94 +/- 0.32% equal error rate on In-The-Wild and 5.07% cross-domain average over four shared datasets, a 28% relative improvement over the best prior frozen-backbone result (Xiao and Vu, 2025) using all 25 layers with identical training data.
Abstract (translated)
URL
https://arxiv.org/abs/2606.30791
PDF
-
MeloDISinger: Melody-Aware & Duration-Preserving Singing Voice Editing with Audio Infilling
2026-06-29 17:22:27Yoonjeong Park, Jaekwon Im, Juhan NamAbstract
Text-based singing voice editing (SVE) aims to revise sung lyrics while preserving the original melody, total duration, and non-edited regions. In this paper, we propose MeloDISinger, a flow-matching-based SVE model for melody-aware and duration-preserving editing. Its core module, MeloDRP, predicts fixed-budget duration ratios, enabling explicit span-wise duration control. For melody-aware duration allocation, MeloDRP fuses phonetic cues with pseudo-MIDI melodic context through cross-attention, while temporal-overlap supervision encourages soft phoneme--note correspondences. We further use a flow-matching mel decoder for audio infilling to synthesize edited regions while preserving surrounding context. In addition, we introduce a duration-aware edited-lyric generation pipeline using WhisperX and an LLM to construct feasible evaluation scenarios. Experiments demonstrate state-of-the-art performance in both objective and subjective evaluations.
Abstract (translated)
URL
https://arxiv.org/abs/2606.30580
PDF
-
SIGMA: Saliency-Guided Sparse Mask Attacks for Speech Emotion Recognition
2026-06-29 16:48:21Qiyang Sun, Yi Chang, Zixing Zhang, Bj\"orn W. SchullerAbstract
Speech conveys rich emotional information. As Speech Emotion Recognition (SER) is usually deployed in privacy-sensitive and reliability-critical environments, adversarial attacks on SER have attracted increasing attention. Existing sparse attacks control the number of perturbed elements, yet, they often lack explainability guidance and explicit measures of explanation consistency. A unified treatment of sparsity and magnitude constraints is also uncommon. In addition, transferability across attack families and target models remains limited. Hence, we propose a SalIency-Guided sparse Mask Attack (SIGMA). On self-supervised speech features, we use post-hoc explainable artificial intelligence (XAI) techniques to produce saliency maps and identify the scope of the mask, and then restrict magnitude-bounded updates to this mask. The mask is computed once and can be reused across models and different sparsity attacks to amortise cost. We evaluate on the IEMOCAP and TESS datasets. Under matched budgets and across multiple sparse-attack settings, SIGMA maintains competitive attack success rates, navigating a conscious trade-off between attack efficacy and explanation consistency. SIGMA therefore provides an efficient and interpretable framework for analysing the vulnerability and explanation behaviour of SER models under structured perturbations.
Abstract (translated)
URL
https://arxiv.org/abs/2606.30550
PDF
-
Predicting Timbre Traits for Interpretable Assessment of Musical Sound Synthesizers
2026-06-29 14:33:44Th\'eo Chasle Cauchy, Modan Tailleur, Lindsey Reymore, Fanny Roche, Mathieu LagrangeAbstract
Measuring neural audio synthesizers' performance is now routinely conducted using distribution based metrics such as the Fréchet Audio Distance (FAD). Although this metric can be correlated with human perception, it offers limited interpretability beyond ranking different approaches. In this paper, we introduce a deep neural timbre trait predictor composed of a pretrained audio neural embedding (CLAP), and a shallow learnable component. The latter is trained using the RWC musical instrument database and human judgments of 20 timbre descriptions (e.g., woody, percussive, rumbling, etc.) for 31 instruments. The resulting model shows strong correlation with average human ratings (r = 0.66, p < 0.001). We then demonstrate the benefit of this predictor for evaluating the performance of TokenSynth, a neural sound synthesizer. First, the Mean Absolute Error (MAE) computed over the set of generated sounds under different conditioning modalities of the model provides the same ranking as a FAD computed with the RWC database as a reference, suggesting that the proposed predictors are able to provide equivalent information on a distributional basis. Second, because the model is able to qualitatively analyze isolated sounds, we can determine which generated sounds could be improved and identify specific timbral dimensions that need adjustment.
Abstract (translated)
URL
https://arxiv.org/abs/2606.30369
PDF
-
EchoHawk: A Reproducible Acoustic Pipeline for Drone Detection, Classification, and Direction-Finding, with a Cautionary Study of Session-Level Data Leakage
Abstract
Passive acoustic sensing is an attractive modality for counter-unmanned aerial system (counter-UAS) defence: it is covert, low-cost, and effective against drones with small radar cross-sections or minimal radio emissions. We present EchoHawk, an open and fully reproducible reference pipeline that detects a drone from its rotor harmonics, estimates its blade-passing frequency, and localises it with a microphone array via classical wideband beamforming (delay-and-sum, MVDR, MUSIC) and time-delay processing (GCC-PHAT, SRP-PHAT), followed by temporal tracking. We evaluate the system on a physically transparent synthetic benchmark that pits drones against hard low-frequency harmonic confusers, such as ground vehicles, and on real recorded audio. Our central methodological contribution is a documented case of session-level data leakage in a widely used public dataset: because its recordings are pre-segmented into short clips, naive clip-level splits place adjacent slices of the same continuous recording in both training and test sets, inflating reported performance. Enforcing recording-session-grouped cross-validation reduces, for example, a random-forest baseline's detection probability at a 1% false-alarm rate from 0.796 to 0.745, yielding honest numbers. All code, figures, and a synthetic data generator are released so that every result runs without any download.
Abstract (translated)
URL
https://arxiv.org/abs/2606.29589
PDF
-
Position-Aware Target Speaker Extraction for Long-Form Multi-Party Conversations: A Diarization-Free Framework for ASR
2026-06-28 16:52:17Yichi Wang, Junzhe Chen, Wangjin Zhou, Tatsuya KawaharaAbstract
In long-form multi-party conversations, highly imbalanced speaker activity and frequent overlap make it difficult to identify "who spoke when and what". Sliding-window continuous speech separation (CSS) mitigates sparse supervision, but often suffers from cross-window speaker inconsistency and residual crosstalk, which in practice requires diarization for reliable speaker attribution. Motivated by the stability of speakers' directions of arrival (DOAs) in meetings, we propose PATSE, a multi-channel Position-Aware Target Speaker Extraction front-end that uses DOA as a spatial prior to directly extract the speech of each target speaker. PATSE combines a DOA-guided spatial encoder and conditioner to generate speaker-attributed streams, from which speaker activity can be inferred via simple post-processing (e.g., VAD) without explicit diarization. Experiments on both replayed and real conversations show consistent ASR gains outperforming CSS and diarization-based pipelines.
Abstract (translated)
URL
https://arxiv.org/abs/2606.29497
PDF
-
DTM-Codec: Dynamic Token Masking for VFR Speech Coding with Efficient Boundary Selection
Abstract
Variable frame rate (VFR) coding has recently emerged in neural speech codecs, allocating fewer frames to redundant regions and more frames to rapidly changing speech. VFR must transmit side information about retained time steps, but prior gains are either not rigorously addressed or often minor once these overhead bits are included in total bitrate. We present Dynamic Token Masking (DTM)-Codec, a neural speech codec that demonstrates clear gains over fixed-frame-rate baselines under a strict matched-total-bitrate protocol. DTM keeps selected encoder tokens, fills masked positions with a learned <MASK> embedding, and transmits a binary keep-mask for position-aware decoding. We further introduce Path Length Equalization (PLE), a linear-time boundary selector for VFR coding that yields well-spread adaptive segments with negligible overhead. Across operating points, DTM-Codec broadly improves reconstruction quality and intelligibility over fixed-frame-rate baselines.
Abstract (translated)
URL
https://arxiv.org/abs/2606.29480
PDF
-
An Optimal Contact-Mechanically Consistent and Flow-Separation Adapted Modeling of Vocal Fold Dynamics
2026-06-27 20:13:31Sardar Nafis Bin Ali, Maryam Naghibolhosseini, Mohsen ZayernouriAbstract
Single mass-spring-damper models of vocal folds have been effective in simulating vocal fold vibrations without added complexity. However, single-degree-of-freedom models cannot sustain oscillation in the presence of structural damping unless source-tract interaction is considered. Moreover, existing lumped models struggle to accurately simulate vocal fold closure during phonation. This study aims to develop a reliable and simplified single-degree-of-freedom model of phonation that can simulate sustained oscillation in a damped system without incorporating a vocal tract model. Additionally, the proposed model maintains vocal fold closure in a manner consistent with the physics of phonation, addressing a longstanding challenge in existing lumped models. High-speed videoendoscopy (HSV) data from four normophonic subjects producing sustained vowel /i/ were used to extract glottal area waveforms (GAWs) via deep learning-based image segmentation for particle swarm optimization of the model parameters. An additional resistance force was incorporated to compensate for flow separation and generate the force imbalance required for sustained oscillation. An external structural force was also added during closure to sustain the closed phase. The 4th-order Runge-Kutta method was used to solve the governing equations with enhanced numerical stability and accuracy. The model parameters were optimized for individual subjects, resulting in normalized errors below 3% between experimental and simulated GAWs. The proposed model accurately reproduced subject-specific vocal fold vibrations and vocal fold closure in agreement with experimental data. Overall, the proposed model provides a computationally efficient framework for simulating sustained phonation without requiring complex source-tract coupling while capturing the key biomechanical and aerodynamic mechanisms of phonation.
Abstract (translated)
URL
https://arxiv.org/abs/2606.29071
PDF
-
Underwater Source Detection and Classification for Signal-based Surveillance: Audio Dataset Curation and Cross-Domain Evaluation
2026-06-27 15:45:54Quoc Thinh Vo, David K. HanAbstract
Machine learning for underwater acoustics is constrained by the scarcity of publicly available labeled datasets. In contrast to air-acoustic domains, where large benchmarks enable rapid model development, underwater datasets are typically small and limited in acoustic diversity, restricting robust model training and cross-domain generalization. To help address this gap, we introduce a curated underwater audio dataset derived from an open-source maritime sound archive. The dataset contains over one thousand labeled audio segments across eight biologically and mechanically relevant acoustic classes, providing an additional resource for training models in data-limited underwater environments. Additionally, we establish a lightweight Convolutional Neural Network (CNN) baseline and propose a margin-enhanced loss with feature alignment to mitigate class confusion arising from data imbalance, acoustic similarity, and cross-domain mismatch. While the baseline achieves 96.35% in-domain accuracy, evaluation on ShipsEar reveals substantial domain shift; the proposed feature alignment improve zero-shot ship detection by 42.60%, demonstrating stronger robustness under distribution mismatch. We further release a transparent curation pipeline and reproducible benchmark to support future research on imbalance mitigation, domain adaptation, and data-efficient underwater acoustic classification.
Abstract (translated)
URL
https://arxiv.org/abs/2606.28988
PDF
-
A Flexible Encoding Model for Non-Unique Note Alignments
2026-06-26 12:38:03Suhit Chiruthapudi, Adam \v{S}tefunko, Silvan Peter, Patricia Hu, Jan Haji\v{c} jr., Carlos Eduardo Cancino-Chac\'onAbstract
Symbolic music alignment links notes in a symbolic performance to their counterparts in a score. While existing alignment encoding formats provide unique correspondences between these notes, there are various musical practices and forms such as practice repetitions in rehearsal and improvised realizations in basso continuo that require a more flexible approach to encoding their alignments. In this paper, we propose a minimal, backward-compatible extension to the Match file format to support such non-unique and semantically complex alignments. We introduce two virtual pointer notes - virtual score notes and virtual performance notes - which allow to encode multiple links between performance and score notes. In addition we expand the Match file's 'section' line to include semantically meaningful annotations of performance regions beyond score-indicated musical repetitions. We further demonstrate the utility of these extensions through two representative use-cases in piano rehearsal and basso continuo.
Abstract (translated)
URL
https://arxiv.org/abs/2606.28032
PDF
-
Grammar-Guided Hierarchical Parsing for Long-form Audio Activity Recognition
2026-06-26 11:10:23Peng Zhang, Qingyu Luo, Philip J. B. Jackson, Wenwu WangAbstract
Long-form audio exhibits an inherent hierarchy: fine-grained events form sub-activities, which in turn constitute higher-level activities. Prior work often models these levels separately, leading to cross-level inconsistencies and requiring supervision at multiple levels. We formulate the problem as hierarchical parsing from event-level evidence: given detected event segments with class posteriors, we infer an order-consistent Act-Sub-Event parse tree. We propose Hierarchical Activity Grammar, encoding hierarchical composition and temporal-order constraints, and perform grammar-guided decoding that combines event evidence with a grammar prior. This yields a temporally grounded parse tree from which sub-activity segmentation and activity classification are derived, without requiring sub-activity or activity labels for training. Experiments on the long-form MultiAct audio dataset demonstrate improved temporal-order consistency (Edit score) and produces interpretable hierarchies.
Abstract (translated)
URL
https://arxiv.org/abs/2606.27965
PDF
-
Advancing Speaker-Based Vocal Effort Classification with WavLM and Data Augmentation in Naturalistic Non-Calibrated Speech Recordings
2026-06-25 20:46:48Zahra Omidi, John H. L. HansenAbstract
The variations in vocal effort range (e.g. whisper, soft, neutral, loud, shout) alter production and speech acoustics, reducing intelligibility and limiting the robustness of any subsequent speech technology. Classification is challenging since effort lies on a continuum, adjacent categories are easily confused, and labeled data remain scarce. Prior SSL approaches with wav2vec2, HuBERT, and AST improve performance on the AVID corpus but still suffer from boundary errors. In this study, we introduce WavLM for the first time in vocal effort classification and benchmark it against wav2vec2 and HuBERT. To address data scarcity, we conduct a systematic study of augmentation strategies, covering RIR convolution, additive noise, time masking, speed perturbation, band-limiting, MixUp, and CutMix. Augmentation consistently improves WavLM, with gains ranging from +0.6% to +1.8% absolute. We further propose Gaussian-neighbor soft labels, which further reduce near-boundary confusions by modeling the vocal effort continuum. Our best system, WavLM-BASE with gradual unfreezing, augmentation, and Gaussian-neighbor soft labels, achieves 78.2% mean accuracy, establishing a new state-of-the-art on AVID.
Abstract (translated)
URL
https://arxiv.org/abs/2606.27543
PDF
-
Learning from Annotation Uncertainty: Entropy-Aware Curriculum for Speech Emotion Recognition
Abstract
Speech emotion recognition (SER) often relies on hard consensus labels that collapse annotator disagreement. We study distribution-based supervision for 9-class SER on MSP-Podcast 2.0 using a WavLM-Base multitask model for categorical emotion and dimensional VAD. Hard-label training is compared with targets from primary and merged primary--secondary annotator vote distributions. Distributional objectives improve alignment with human vote distributions, reducing JSD/KLD relative to hard-label training. Analysis shows that hard supervision partly benefits from assigning ambiguous utterances to the residual Other class, whereas distributional supervision redistributes uncertainty across emotion categories. Entropy-stratified evaluation shows that high-ambiguity utterances remain challenging, but distribution-based supervision better captures perceptual uncertainty. These findings support moving beyond hard labels toward targets that reflect listener disagreement.
Abstract (translated)
URL
https://arxiv.org/abs/2606.27536
PDF
-
Elastic Time: Dynamic Frame Rate Bottlenecks for Neural Audio Coding
2026-06-25 17:33:09Dimitrios Bralios, Paris Smaragdis, Minje KimAbstract
Neural audio autoencoders have become a core component of compression, feature extraction, and generation. However, while existing systems support variable bitrate, the vast majority of models still operate at a fixed latent frame-rate, allocating equal temporal budget to regions with very different information density, which can result in unnecessarily long sequences. We introduce Elastic Time, a dynamic frame-rate bottleneck that converts fixed-frame-rate autoencoders to dynamic ones. Our method learns a lightweight latent predictor used to decide which frames can be skipped and later reconstructed, enabling efficient greedy boundary selection at inference. Experiments show our method enables deployment-time rate control while improving efficiency-quality tradeoffs relative to baselines. Overall, we provide a flexible mechanism for adjusting temporal resolution in audio autoencoders, potentially facilitating more efficient downstream modeling for generation and long-context tasks.
Abstract (translated)
URL
https://arxiv.org/abs/2606.27320
PDF
-
voxmap-studio: An open-source speaker diarization annotation tool with built-in cost instrumentation
Abstract
Labeling speaker diarization data is costly, yet annotation tools rarely measure that cost. We present voxmap-studio, an open-source, React-based diarization annotation tool integrated with the pyannote-based diarization ecosystem. Its canvas is initialized by a fast stride-accelerated diarization engine so that the annotator corrects a hypothesis rather than drawing every speaker turn by hand, and the tool records annotation cost - typed edit-operation counts and time - as a first-class output, enabling quantitative comparison of how much different forms of assistance actually help. Export is gated on per-segment human confirmation and guarded by injected "phantom" attention checks, which prevent unverified automatic output from being released as ground truth. In a preliminary study on nine AMI audio files, unassisted manual annotation was the costliest and least accurate, and automatic initialization shifted the work from creating turns to correcting them; highlighting uncertain segments gave the lowest cost in our small sample. The tool and its instrumentation are open source.
Abstract (translated)
URL
https://arxiv.org/abs/2606.26842
PDF
-
wav2tok 2.0: Scalable Audio Tokenization Maintaining Explicit Pairwise Token Alignment for Efficient Audio Retrieval
2026-06-25 10:04:35Adhiraj Banerjee, Vipul AroraAbstract
Learning discrete speech representations that preserve similarity across variable-length utterances is central to query-by-example spoken term detection (QbE-STD). While wav2tok introduced CTC-based sequence alignment to enforce token consistency, its tightly coupled clustering and alignment training recipe limits scalability. We propose wav2tok 2.0, a scalable alignment-aware speech tokenizer built on the BEST-STD backbone. wav2tok 2.0 employs staged training, first learning discriminative, speaker-invariant representations via contrastive learning and vector quantization, and then enforcing pairwise token consistency using a CTC alignment loss and a novel DTW-aligned framewise prediction objective with adaptive weighting. Experiments show that wav2tok 2.0 consistently outperforms BEST-STD and general-purpose tokenizers on QbE-STD while remaining efficient and scalable.
Abstract (translated)
URL
https://arxiv.org/abs/2606.26824
PDF
-
WQ-Fusion: Dynamic Gated Attention for Cross-Domain Audio Representation
2026-06-25 03:07:34Mingda Lin, Lei Ding, Xinyue Zhou, Tiantian Xiong, Hanchen Pei, Gongping Huang, Hao Zhang, Jingdong Chen, Jacob BenestyAbstract
While pre-trained models excel in specialized tasks, learning universal representations across diverse acoustic domains remains challenging. To address this, we propose WQ-Fusion, a robust dual-encoder framework for cross-domain audio representation learning. Overcoming the limitations of static concatenation, WQ-Fusion integrates whisper and qwen via an Adaptive Feature Modulation module and a novel element-wise gated attention mechanism. This design enables dynamic feature selection, allowing the model to selectively emphasize relevant acoustic and semantic dimensions. Extensive experiments on the Interspeech 2026 Audio Encoder Capability Challenge (Track A) benchmark demonstrate that by effectively routing heterogeneous information, WQ-Fusion achieves a superior overall score of 0.836, significantly outperforming the strongest single-encoder baseline.
Abstract (translated)
URL
https://arxiv.org/abs/2606.26556
PDF
