每日 arXiv

音频 · 语音 · 音乐 · 声学

每日新爬取的 arXiv 论文，按评分排序。中文摘要帮你快速判断是否值得深读。

2026-07-15
日期2026-07-15
已评分—
均分—
最高—

Daily Papers — 2026-07-15

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

1. MetaPerch: Learning from metadata for bioacoustics foundation models

Authors: Mustafa Chasmai, Vincent Dumoulin, Jenny Hamer

Categories: cs.LG, cs.SD | Accepted to ICML 26 Score: 6.80/10 (Obj:8 Id:5 Ind:9 Comp:6 Eff:6 Nov:8 Rep:9)

  • Strength: 首次在 bioacoustics foundation model 中系统研究 9 种 per-recording metadata 作为辅助多任务损失，在 17 个数据集上 vs 同架构 baseline 取得 +0.006-0.070 ROC-AUC 提升，在 BEANS 上达到 SOTA（mean acc 0.870, cmAP 0.512），代码开源、5 seeds、超参数完整公开。
  • Weakness: vs BioBaseline 提升幅度偏小（多数 0.006-0.019，部分在 1-2 sigma 内），在 BirdSet 上略输 Perch 2.0（0.906 vs 0.908），未与”加随机辅助任务”最简对照比较，未在完整 Perch 2.0 架构上验证 metadata 增益是否被 source prediction 吸收。
  • Full review: Claude Code 全文七公理审稿

Bioacoustic foundation models rely on large-scale citizen science platforms like Xeno-Canto for geographically and ecologically diverse data. Recent work has shown that supervision alone can produce SotA species detection models when trained on this large-scale data – however, there remains unutilized potential in the form of recording metadata readily available within these community-driven data hubs. In this work, we explore the use of metadata – such as location and time – as auxiliary supervision signals, allowing the model to leverage species-metadata correlations in its learned representation. Auxiliary metadata losses provide additional information beyond vocalizations alone that can encourage a richer, more robust representation that generalizes better to species distribution and acoustic domain shifts – important challenges for deployment in real-world passive acoustic monitoring (PAM) settings. We introduce MetaPerch, a new foundation model that achieves strong species identification performance across multiple challenging domains and present an extensive empirical study of the effects of 9 diverse metadata sources on 17 bioacoustic datasets.

2. Auditing Protocol-Level Shortcuts in Large Audio Language Model Judges for Speech Evaluation

Authors: Joonyong Park, David M. Chan, Yuki Saito, Hiroshi Saruwatari

Categories: cs.SD, cs.CL, eess.AS Score: 6.80/10 (Obj:9 Id:8 Ind:8 Comp:8 Eff:6 Nov:7 Rep:5)

  • Strength: 首次将三种 LALM judge 部署协议统一在”protocol-level shortcut”框架下审计，发现 wrong+audio 条件下 5/6 评委 emotion accuracy ≤0.10（即使音频在场仍复制错误 specialist 标签），并通过 emotion vs. language 跨属性对比识别出 capability-dependent shortcut 这一可迁移解释机制。
  • Weakness: 面板规模有限（6 评委 × 4 属性，跨属性泛化仅基于 2 个数据点），代码与完整 prompt 未开源，闭源评委 API 版本不可控，且论文未提供任何修复方向。
  • Full review: Claude Code 全文七公理审稿

Large audio-language models (LALMs) are increasingly used as automatic judges for speech evaluation. However, high agreement with human ratings does not guarantee that their verdicts are grounded in the audio. A judge may instead rely on specialist labels or reference data supplied by the evaluation protocol itself, taking a shortcut in place of listening to the audio. In this paper, we audit such protocol-level ``shortcuts’’ in LALM judges across three common deployment protocols: feature-blueprint judging, where the audio is replaced by a structured text description of acoustic features, reference-conditioned judging, and pairwise A/B comparison. Across six judges and four attributes, we find that several LALMs rely on protocol-level shortcuts. For example, in feature-blueprint judging, incorrect specialist labels reduce five judges’ emotion accuracy to 0.10 or below, and in concatenated A/B comparisons, Qwen3-Omni-Thinking often picks the same slot regardless of order swaps. These results indicate that aggregate agreement can overstate the validity of LALM judges unless the model and the evaluation protocol are assessed jointly, and that each model-protocol pair should be evaluated with a matched shortcut probe.

3. Cover First, Disagree Softly: Rethinking Mismatch-First Active Learning for Frame-Level Audio Classification

Authors: Shiqi Zhang, Tuomas Virtanen

Categories: eess.AS, cs.AI, cs.SD | submitted to DCASE workshop 2026, under reviewing Score: 6.68/10 (Obj:8 Id:9 Ind:8 Comp:8 Eff:6 Nov:6 Rep:5)

  • Strength: 通过严格的 2×3 因子分解实验（2 个几何骨干 × 3 种分歧使用方式 × 2 个数据集 × 10 个种子），清晰诊断了 MFFT 在低预算下的两个失败症状（群组冗余 + 离群点漂移），并提出无超参数的 MW-FL 方法，在两个数据集上均获最佳 AULC (p≤0.006)，对 mismatch-first 家族的提升达 +0.031~+0.046。
  • Weakness: 对最强基线 plain FL 的提升仅 +0.004 AULC（虽统计显著但实际意义有限），仅测试 2 个数据集且未与次模 AL 通用方法（SIMILAR, BatchBALD, BADGE）比较，无代码发布，泛化证据基础偏窄。
  • Full review: Claude Code 全文七公理审稿

Sound event detection relies on frame-level strong labels whose annotation is expensive. Active learning addresses this problem by selecting the audio segments whose labels help the classifier most. One of the prevailing acquisition strategies for this task, mismatch-first farthest-traversal (MFFT), combines the disagreement between two classifiers and the diversity of the selected segments through hard sequential decisions. It selects whole groups of high-disagreement segments first and spreads only the remaining budget by farthest traversal. On two multi-label datasets we show that this design is blind to the similarity among the selected segments and fails under low budgets, with every mismatch-first variant ending below the plain geometric strategy it builds on. We propose mismatch-weighted facility location (MW-FL), which spends the entire budget through a disagreement-weighted coverage objective that penalizes similarity among the selected segments. The disagreement signal from MFFT is used to obtain the nonnegative weights of this facility-location objective, without introducing hyperparameters. Experiments across two geometric mechanisms with three ways of using disagreement show that coverage of the selected segments is the dominant factor, hard disagreement gating of selection is harmful on both mechanisms, and soft disagreement weighting helps on top of coverage. MW-FL attains the best area under the learning curve on both datasets.

4. Greedy Volume Maximization of Gradient Embeddings for Long-Tailed Frame-Level Bioacoustic Active Learning

Authors: Shiqi Zhang, Marius Faiß, Ariana Strandburg-Peshkin, Tuomas Virtanen

Categories: eess.AS, cs.AI, cs.SD Score: 6.60/10 (Obj:9 Id:6 Ind:9 Comp:6 Eff:8 Nov:6 Rep:5)

  • Strength: BADGE-Greedy-DPP 在 HyenaSET 10 类帧级任务上 10 次运行 N-AULC 56.7%（比最强基线 MFFT +1.2pp，Holm 校正 p=1.60e-3），Rare-N-AULC +3.8pp，是唯一在 10/10 次运行中达到全监督参考水平的方法，预算仅用 3.94% 训练集；贪心 log-det 遍历提供 (1-1/e) 近似保证。
  • Weakness: 帧级残差加权聚合（第二贡献）无独立消融（所有 BADGE 变体共用该构造，未与段级 BADGE 对比），且仅在单一未公开数据集 HyenaSET 上验证，代码未发布。
  • Full review: Claude Code 全文七公理审稿

Bioacoustic call-type classification relies on costly expert annotation. Active learning can reduce this burden by selecting a small batch of segments for expert annotation and using the labeled segments for training the classifier. The setting is hard: the target calls are extremely sparse and the call-type distribution is long-tailed, so a tight budget must be spent on the few rare, informative segments. We propose BADGE-Greedy-DPP, a deterministic batch selector that greedily adds the segment whose BADGE gradient embedding most enlarges the volume spanned by the batch; because this log-volume objective is submodular, the greedy rule guarantees a batch value at least a (1-1/e) fraction of the optimum of this objective, a guarantee not provided by BADGE’s existing k-means++ and MCMC DPP sampling heuristics. There is also a temporal granularity mismatch in the task. The acquisition function scores whole segments, yet the informative frames inside them are few. Uniform averaging therefore washes them out. We show that the BADGE construction naturally addresses this mismatch when applied frame-wise, as prediction residuals weight the aggregated pseudo-gradient, so confidently predicted no-call frames contribute little while a single uncertain rare-call frame can still set the segment’s direction. Across 10 runs on a sparse, imbalanced hyena call-type dataset, BADGE-Greedy-DPP achieves the best overall and rare-call-type performance among all compared query strategies, including MFFT, the strongest non-BADGE baseline, and the two vanilla BADGE traversals.

5. Improving Text-to-Audio Instruction Following via Fine-Grained Feedback from Audio-Aware Large Language Models

Authors: Chun-Yi Kuan, Siwon Kim, Byeonggeun Kim, Suyoun Kim, Bo-Ru Lu et al.

Categories: eess.AS, cs.AI, cs.CL, cs.LG, cs.SD | Accepted to the Long Paper Track at Interspeech 2026 Score: 6.30/10 (Obj:8 Id:6 Ind:5 Comp:6 Eff:8 Nov:5 Rep:6)

  • Strength: AJPO 框架在 MultiEvent-Temporal-3 上将 Joint Accuracy 从 baseline 32.8% 提升至 55.0%（+22.2pp），在 S3Bench 上达到 49.9%（vs. CRPO 45.4%），且标准音频质量指标未退化，消融实验系统验证了 fine-grained ALLM feedback 相对 CLAP-based similarity 的必要性。
  • Weakness: 训练信号（Qwen2.5-Omni-3B 构造 preference）与评测指标（Qwen2.5-Omni-7B 计算）来自同一模型家族且使用相同评判协议，存在 major 级别循环论证风险；方法仅在一个 backbone（TangoFlux）上验证，未排除方法对特定架构的依赖性。
  • Full review: Claude Code 全文七公理审稿

Recent text-to-audio models generate high-quality audio, but often fail to follow instructions involving multiple sound events and temporal order. This gap arises because existing evaluation and training signals mainly emphasize global similarity or perceptual quality, with limited supervision on instruction-level correctness. We propose an instruction-level framework that uses audio-aware large language models (ALLMs) as fine-grained judges to verify target event presence and temporal relations in generated audio. After validating ALLM judgments on benchmarks and through human verification, we use their feedback to construct preference pairs for direct preference optimization. We further introduce S3Bench, a narrative benchmark for evaluating multi-event temporal instruction following. Experiments show that our method improves event completeness, temporal ordering, and joint instruction-following accuracy across existing benchmarks and S3Bench, while maintaining audio quality.

6. From Prediction to Collaboration: Interactive Symbolic Music Analysis

Authors: Emmanouil Karystinaios, Johannes Hentschel, Markus Neuwirth, Gerhard Widmer

Categories: cs.SD, cs.AI | in Proceedings of the 27th International Society for Music Information Retrieval Conference (ISMIR) 2026 Score: 6.26/10 (Obj:8 Id:5 Ind:8 Comp:5 Eff:6 Nov:5 Rep:9)

  • Strength: 首次将符号 RN 分析重构为”预测→分析连续体”，在 Dilemmadata 的 DLC 子集上 RN accuracy 达 .578（vs AnalysisGNN .516，+0.062），掩码补全从 5% 已知标签 .577 单调上升至 95% .831，代码/数据/模型全开源。
  • Weakness: 核心架构是作者两篇前期工作（RNBert + AnalysisGNN）的直接组合，无新机制；盲推理在 AugNet 上 vs RNBert 仅 +.002（持平）；关键消融缺失（MusicBERT vs 图模块 vs 混合三路隔离、掩码模型 preservation losses 分解）；遗漏 frog（NADE-style）基线；未做用户研究验证交互效用。
  • Full review: Claude Code 全文七公理审稿

Automatic symbolic music analysis has made substantial progress, yet existing systems are typically designed for a single mode of use, such as full-score prediction, and therefore do not match the broader range of operations that arise in analysis workflows, including partial completion, local correction, and iterative refinement. As a result, there remains a gap between strong benchmark models and systems that can support interactive analytical use. We present a unified framework for symbolic Roman-numeral (RN) analysis that narrows this gap by combining strong predictive performance with direct support for constrained completion and revision. The method is designed to provide a practical trade-off between accuracy and interactive responsiveness by computing expensive pretrained representations once and reusing them during iterative refinement, making powerful pretrained models more amenable to interactive settings. It supports complete score analysis, targeted revision of existing labels, and inference of missing annotations from partial context through a shared modeling framework. Experiments on Dilemmadata, the largest and most heterogeneous benchmark of its kind, show that the proposed approach is a strong RN-analysis baseline while also supporting masked completion from partial labels. Together with a prototype interface for multi-level candidate inspection and editing, these results position automatic RN analysis not only as a prediction problem, but also as a foundation for future interactive tools for music analysis.

7. Self-supervised Speech Comparison for L2 Phone, Rhythm, and Intonation Scoring

Authors: Stephen McIntosh, Reuben Smit, Daisuke Saito, Nobuaki Minematsu, Herman Kamper

Categories: cs.CL, eess.AS Score: 5.95/10 (Obj:8 Id:5 Ind:8 Comp:6 Eff:5 Nov:6 Rep:6)

  • Strength: DTW warp path 的 tempo irregularity 和 interval distortion 两个 rhythm 指标真正新颖，在 English sentence rhythm 上接近人类 inter-rater 水平 (r=44.9 vs human 45.3)，且 phonetic sentence scoring 显著超过人类 (r=57.6 vs 53.6)。
  • Weakness: Intonation scoring 效果明显不足 (r=32.9 vs human 45.3)，且缺少与 2023 Indicon 同类工作 (SSL+DTW for unsupervised pronunciation assessment) 和 supervised SOTA (NOCASA 2025) 的比较，k-means residual 方法已有 2025 年先前工作 (2502.19387) 降低了 intonation。
  • Full review: Claude Code 全文七公理审稿

L2 speech assessment has traditionally focused on phonetic assessment, leaving the scoring of suprasegmental features such as rhythm and intonation underexplored. Moreover, assessment methods often require training with labeled L2 speech data, making them difficult to apply in low-resource settings. We investigate whether DTW over self-supervised WavLM representations can provide a text-free framework for assessing phonetic accuracy, rhythm, and intonation in English and Japanese L2 speech. Results show that a basic DTW-based approach that compares learner speech to native templates exceeds human agreement on holistic and sentence-level phonetic scoring. For rhythm, we introduce methods that measure the degree of warping in the DTW alignment path; our best method approaches human-level performance. For intonation, we combine DTW distance over prosodic residuals with pitch and intensity features, but performance remains more modest on some tasks. Our results point to self-supervised representations as a promising, text-free basis for multi-aspect pronunciation assessment.

8. Rethinking Speech Foundation Model Fine-tuning: Better SFT or Better Match?

Authors: Wangjin Zhou, Yizhou Zhang, Yichi Wang, Tatsuya Kawahara

Categories: cs.SD | Accept by Interspeech 2026 Score: 5.50/10 (Obj:5 Id:5 Ind:9 Comp:4 Eff:5 Nov:5 Rep:6)

  • Strength: 在 9 个公开 SSL checkpoint × 8 种 SFT 配置 × 3 个 SUPERB 任务上系统验证了”top-group SFT recipe 的身份依赖于预训练实例”，用 paired McNemar 检验和多种子实验提供了统计证据，种子翻转现象直接支持激活可靠性解释。
  • Weakness: 核心概念”elicitation match”缺乏可操作化度量（无可计算指标预测 recipe-checkpoint 匹配度），仅覆盖 3 个分类任务和 2 个配置轴，大量 convergence anomaly 未被机制性解释，且未与 SUPERB leaderboard SOTA 或 NLP 领域同类工作（Merchant et al. 2020）深入对比。
  • Full review: Claude Code 全文七公理审稿

Supervised fine-tuning (SFT) is widely used to adapt self-supervised speech representations to downstream classification tasks. Small gains observed under a single pretrained checkpoint are often interpreted as method-level improvements, i.e., a higher attainable performance ceiling. We show that such conclusions are not always reliable because SFT outcomes depend strongly on the specific pretrained instance. We conduct a systematic study on 3 SUPERB classification tasks, evaluating 8 SFT variants across 9 pretrained checkpoints from wav2vec~2.0, HuBERT, and WavLM, with multi-seed repetitions on representative base-scale models. We find that the identity of the statistically indistinguishable top-group SFT recipe is often checkpoint-dependent, with limited transferability across pretrained instances. These findings suggest that many reported downstream gains reflect instance and seed dependent elicitation match, rather than universally improving the attainable performance ceiling.

9. Audio-Text Cross-Attention with Psycholinguistic Support Features for Ambivalence/Hesitancy Recognition

Authors: Luiz F. B. F. Martins, Rodrigo W. Pisaia, Matheus M. Girardi, Isabella Berkembrock, João A. Almeida et al.

Categories: cs.CV Score: 5.30/10 (Obj:8 Id:5 Ind:5 Comp:4 Eff:5 Nov:4 Rep:8)

  • Strength: 在 BAH 数据集上提出 psycholinguistic-grounded 74 维 support features 注入 cross-attention + MIL 架构，development set AP=0.875，优于 mean pooling baseline (+0.047 AP)，代码开源。
  • Weakness: 消融不充分（cross-attention/ensemble/injection 位置均未单独验证），仅报 development set 结果无私测，同期竞赛工作 BROTHER (F1=0.7465) 和 Simple Features (F1=0.731) 均优于本文 (F1=0.722)，核心 novelty 实质是已有 psycholinguistic 工具的工程组合。
  • Full review: Claude Code 全文七公理审稿

We present an audio-text system for the Ambivalence/Hesitancy Video Recognition Challenge of the 11th ABAW Competition. The method excludes visual frames and represents each video as overlapping 5-second windows aligned with transcript timestamps. Each window combines a 320-dimensional prosodic audio descriptor, a 768-dimensional emotion-oriented RoBERTa embedding, and 74 handcrafted features capturing uncertainty, hedging, and attitudinal conflict. Audio and text are fused via temporal cross-attention, while support features are injected prior to gated multiple-instance learning (MIL) pooling to modulate the window’s importance. Predictions from five independently initialized models are averaged. On the labeled public development set, the ensemble achieved an average precision of 0.875 and a macro-F1 of 0.72. Our source code is publicly available at https://github.com/Liga-de-IA-PUCPR/abaw-11-ah-challenge/.

10. Music-to-Dance Generation via Atomic Movements

Authors: Xinhao Cai, Yixuan Sun, Minghang Zheng, Qingchao Chen, Xin Jin et al.

Categories: cs.CV, cs.AI Score: 5.26/10 (Obj:6 Id:5 Ind:5 Comp:5 Eff:6 Nov:5 Rep:6)

  • Strength: 提出 LLM-assisted atomic movement 标注 pipeline 和 D3PM frame-level discrete diffusion planner，在 AIST++ 上 FIDk=25.26、BAS=0.2470 均优于 EDGE/Lodge/Bailando，RR-precision 26.6 显著高于所有 baseline，验证了结构化规划对舞蹈生成的增益。
  • Weakness: 核心标注 pipeline 依赖闭源 Gemini-2.5-Pro 且无人类标注校验，baseline 比较遗漏 Lodge++(TPAMI 2025) 和 DanceMeld(2024) 等直接竞争方法，仅在单一数据集 AIST++ 上测试且无人类评估验证”perceptual naturalness”声称。
  • Full review: Claude Code 全文七公理审稿

Music-driven dance generation aims to produce human motion that is both rhythmically synchronized and semantically consistent with music. While recent neural approaches have achieved impressive visual realism, they typically model motion as a continuous signal and neglect its compositional nature, making generated dances structurally incoherent and difficult to control. In this work, we introduce a structure-aware framework that models choreography as a sequence of atomic movements-semantically interpretable motion events that serve as the building blocks of dance. To construct this atomic movement vocabulary, we first segment large-scale dance data and cluster them into atomic movement groups. We then employ a large language model to semantically relabel and refine the clusters, yielding a set of interpretable and reusable atomic movements. Based on these atomic movement annotations, we design a two-stage generation framework that mirrors the human choreography process. In the atomic movement planning stage, the model predicts the type, duration, and timing of atomic movements conditioned on the input music, forming a symbolic dance allocation. In the completion stage, a transition-aware generator synthesizes smooth and stylistically coherent motion conditioned on the planned structure. Extensive experiments demonstrate that our method produces dances with significantly improved structural coherence, rhythmic alignment, and perceptual naturalness compared to existing baselines, while providing enhanced interpretability and controllable editing through explicit structural representation.

11. Genre Bias or Aesthetic Perception? Identifying and Mitigating Shortcut Learning in Music Evaluation

Authors: Yizhou Zhang, Wangjin Zhou, Yi Zhao, Wei Tan, Keisuke Imoto et al.

Categories: cs.SD | Accept by ISMIR 2026 Score: 4.95/10 (Obj:6 Id:5 Ind:5 Comp:4 Eff:5 Nov:5 Rep:6)

  • Strength: 系统诊断揭示 SongEval 的预测分与 pop-proximity 方向强相关（Spearman 0.81-0.88），在 genre-balanced 控制实验中仍存在显著 genre effect（ANOVA p<0.001），为音乐评估模型的 genre shortcut 提供了可靠经验证据。
  • Weakness: 方法是 focal regression loss + group EMA regularization 的直接组合，两个组件均来自成熟技术（Focal Loss 2017, group fairness regularization），提升幅度小（overall accuracy +2.8pp），residual pop-proximity SRCC 仍高（0.62-0.68），且未与 group DRO 等最直接 baseline。
  • Full review: Claude Code 全文七公理审稿

Music aesthetics scoring plays a critical role in applications such as dataset curation, generative model evaluation, and reward modeling for music generation. Recent approaches rely on deep neural networks trained on human-annotated ratings, but these models may exploit spurious correlations rather than capturing perceptually meaningful aesthetics. In this work, we identify a previously underexplored failure mode in music evaluation models: genre-induced shortcut learning. Through a systematic analysis of SongEval, we show that biases in training data lead to strong correlations between genre-related features and predicted scores, causing the model to use them as a proxy for aesthetics. This results in systematic overestimation of pop music and undervaluation of high-quality samples from other genres, leading to predictions that are inconsistent with human preferences. To address this issue, we propose a training objective that jointly reweights hard samples and regularizes group-level performance, encouraging the model to learn genre-invariant representations of musicality. Experimental results demonstrate that our method reduces genre-dependent bias and improves alignment with human preferences, as reflected by gains in both cross-genre and within-genre preference alignment.

12. Learning Speaker Identity Beyond Language and Modality Constraints: Insights from the POLY-SIM 2026 Challenge

Authors: Marta Moscati, Muhammad Saad Saeed, Marina Zanoni, Mubashir Noman, Rohan Kumar Das et al.

Categories: cs.CV | Accepted at ACM MM 2026 Score: 4.95/10 (Obj:6 Id:5 Ind:5 Comp:4 Eff:5 Nov:4 Rep:8)

  • Strength: 组织了 18 团队参与的标准化挑战赛，top 方法将 baseline 从 73.37% 提升至 99.89%，并识别出 test-time unsupervised adaptation 是性能提升的关键驱动因素。
  • Weakness: 仅覆盖 English-Urdu 一对语言且 top-3 均超 99%（benchmark 趋于饱和），分析深度不足（Lessons Learned 仅 2 点操作层面观察），且 top 方法利用测试时输入数据做无监督适应的做法在真实部署中不可行，评测-部署 gap 未被独立验证。
  • Full review: Claude Code 全文七公理审稿

Multimodal speaker identification systems typically assume the availability of complete and homogeneous audio-visual modalities during both training and testing, and assume each speaker only speaks a single language. However, in real-world applications, such assumptions often do not hold. Visual or audio information may be missing due to occlusions, camera or microphone failures, or privacy constraints. Multilingual speakers introduce additional complexity due to linguistic variability across languages. These situations constitute substantial challenges for the robustness and generalization capabilities of multimodal speaker identification systems. Aim of the POLY-SIM 2026 challenge is to address these aspects of speaker identification and to provide a standardized setup for the comparison of the proposed solutions.

13. CF-Net: Conflict Fusion with Speaker Normalisation and Certainty Weighting for Ambivalence/Hesitancy Recognition

Authors: Tung Hung Bui, Hong Hai Nguyen, Van Thong Huynh

Categories: cs.CV | 6 pages, 3 figues, 3 tables, 26 references Score: 4.58/10 (Obj:8 Id:5 Ind:6 Comp:4 Eff:5 Nov:2 Rep:6)

  • Strength: 系统性地将 speaker normalisation + certainty-weighted focal loss + ConflictFusion 组合在 frozen-backbone 框架下，val-to-test 正向迁移 (+0.021) 符合 reduced identity overfitting 假设，训练仅需 15 分钟/单 GPU。
  • Weakness: 核心融合机制直接采用 Bekhouche et al. (2026) 的 conflict-aware fusion，其余组件均为成熟标准技术，无新机制或 synergy 证明；val F1 (0.716) 低于已发表 baseline (0.746/0.736)，test F1 跨挑战赛比较不严格，单次 seed 无方差报告。
  • Full review: Claude Code 全文七公理审稿

Detecting ambivalence and hesitancy (AH) in unconstrained video is challenging because the target signal is inherently ambiguous and expressed through subtle cross-modal incongruence rather than prototypical affect. We present CF-Net, a deep multimodal network submitted to the 3rd Edition of the AH Video Recognition Challenge (ABAW 11th, ECCV 2026), targeting the BAH dataset. CF-Net encodes visual, audio, and transcript streams with frozen SigLIP2, HuBERT, and DistilBERT backbones, normalises backbone features per speaker to reduce identity leakage, and fuses them via a ConflictFusion module that explicitly computes pairwise cross-modal incongruence. Training combines certainty-weighted focal loss, manifold mixup, and modality dropout; an auxiliary certainty-regression head leverages ambiguity annotations to stabilise learning on genuinely borderline samples. CF-Net achieves a Macro F1 of 0.7155 on the BAH validation set and 0.7364 (AP = 0.7492) on the private challenge test set.

14. From Continuous Deployment to Queryable Dataset: Terabyte-Scale AIS-Aligned Passive Acoustic Labelling

Authors: Wayne Renaud, Priyanka Aravindan, Gabriel Spadon

Categories: cs.SD, cs.DB | OCEANS’26 - Monterey Score: 3.95/10 (Obj:5 Id:2 Ind:5 Comp:5 Eff:4 Nov:4 Rep:4)

  • Strength: 将标准 PostGIS spatiotemporal join 应用于被动声学存档与 AIS 轨迹对齐，在 Station 17 上验证了 35,679 个窗口的分类和 SNR-vs-distance 物理关系复现，理念上从固定数据集转向可查询数据产品。
  • Weakness: 无定量 baseline 对比（无运行时间/内存/scalability 比较）、无下游 ML 任务验证、未引用已有同类工作（PORTIA Dataset, Belgian North Sea AIS-annotated recordings）、核心工具 AISdb 为共同作者自有且未披露利益冲突、代码和数据未开源、仅用最小站点验证而未展示声称的 terabyte-scale 完整结果。
  • Full review: Claude Code 全文七公理审稿

Long-duration passive acoustic deployments produce large archives of recordings that are not linked to vessel tracks or encounter structure, leaving range and contact conditions unavailable as variables and requiring manual selection for analysis. To address this limitation, we propose a database-native workflow that aligns hydrophone recordings with Automatic Identification System (AIS) position reports to produce distance-resolved data. Fixed-duration recording windows and AIS messages are stored as persistent geospatial tables and associated through an indexed spatiotemporal join, replacing in-memory nested iteration with a single scalable set-based database process capable of handling continuous, multi-year, million-window archival deployments without exhausting available memory. In this study, the approach processes approximately 9.5x10e5 recording windows and 6.9x10e6 AIS position reports, producing a structured table that separates no-contact, single-contact, and two-contact windows, with the closest point of approach computed directly where applicable and background conditions characterized via deterministic spectral ranking. This formulation enables a GeoAI framework in which spatially indexed, queryable data become directly usable for machine learning. The resulting data product reveals predominantly noise-dominated conditions, with vessel contributions emerging mainly at shorter ranges, indicating that the task lies in extracting structure under background-limited regimes. Spectrogram and quantitative analyses show weak tonal signatures embedded in noise and a consistent decay of signal-to-noise ratio with distance, supporting the use of this representation for scalable machine learning, similarity analysis, and predictive acoustic modelling in real maritime environments.

15. Bring Music The Horizon: Music-Driven 360$^\circ$ Video Generation

Authors: Kai Hsu Tsai, Yong Wei Fu, Hung I Yang, Yu-Chih Chen

Categories: cs.CV, cs.MM, cs.SD, eess.AS, eess.IV | 5 pages, 1 figure Score: 2.32/10 (Obj:5 Id:2 Ind:2 Comp:2 Eff:2 Nov:2 Rep:2)

  • Strength: 将音乐情绪轨迹（bar-level V-A 预测）与 360° 全景视频生成结合，pipeline 设计清晰，使用了 EmotiCrafter + SEGA + SDXL 360 LoRA + Wan I2V 等组件，概念上完整可运行。
  • Weakness: 全文 5 页 1 图，无任何定量实验、无消融、无用户研究、无客观指标，唯一自研组件（Dynamic V-A Regressor）完全未描述且未评估，仅与一个非 360° baseline 做定性对比，核心贡献为 6 个已有工具的线性拼装。
  • Full review: Claude Code 全文七公理审稿

Music visualization offers a powerful way to enhance listeners’ understanding and experience of music by translating auditory signals into visual forms. However, most existing approaches either rely heavily on lyrics or generate flat, non-immersive videos similar to conventional music videos, which limits their ability to convey the emotional dynamics of music and provide an immersive listening experience. We propose Bring Music The Horizon, an emotion-aware pipeline for music-driven 360$^\circ$ video generation. Given an input song, our work first estimates its emotional trajectory by predicting valence-arousal values at the level of every four bars. These values are then converted into emotion-aware visual guidance using EmotiCrafter, and these guidance vectors can be manipulated by the SEGA framework, which provides fine-grained semantic control for keyframe generation. Finally, image-to-video models are applied to the generated keyframes to synthesize temporally continuous 360$^\circ$ videos for immersive music visualization. Our pipeline generates 360$^\circ$ music visualization videos that reflect the emotional progression and temporal structure of the input song. We demonstrate its capability using songs from different genres and provide qualitative comparisons with From-Sound-To-Sight, a representative audio-to-visual generation baseline, on our project page at https://etoile-et-toi-mp3.github.io/BMTH_Project_Page/.