Pith. sign in

REVIEW 4 major objections 5 minor 45 references

HyPASE: Hyperbolic Geometry for Parameter-Efficient Speech Emotion Fine-Tuning Framework for Large Audio-Language Models

T0 review · 4 major / 5 minor · reviewed 2026-08-08 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A 0.12%-parameter hyperbolic adapter beats Euclidean PEFT baselines for speech emotion recognition.

desk verdict HyPASE's empirical results are real but the 'hyperbolic adapter' collapses to plain diagonal scaling; the genuine hyperbolic contribution is EMCA's fusion, so the paper needs reframing before its headline claim is taken seriously. read the letter →

arxiv 2608.04351 v1 pith:XE6ZBCEG submitted 2026-08-05 cs.SD cs.AI

Tian Jin , Ruikang Zhang , Zefeng Zhao , Ding Luo , Jin Zeng This is my paper · ORCID
classification cs.SDcs.AI
keywords speechemotionrecognitionlargeaudio-languagemodelsparameter-efficientfine-tuninghyperbolicgeometryPoincaréballmulti-granularityrepresentationclassimbalancezero-shotcross-datasetgeneralization
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper proposes HyPASE, a parameter-efficient fine-tuning framework that adapts a frozen large audio-language model for speech emotion recognition by working in hyperbolic space rather than flat Euclidean space. The central bet is that the hyperbolic radius, the geodesic distance from the center of a Poincaré ball, is a faithful measure of how coarse or fine an emotion cue is, with broad prosodic patterns near the center and fine emotional semantics closer to the boundary. Under that assumption, the framework trains only about 0.12% of the model's parameters, yet outperforms Euclidean PEFT baselines on MELD across all reported metrics and improves minority-class recognition on IEMOCAP. If correct, this would mean that non-Euclidean geometry is a practical, low-cost axis for steering large audio-language models toward tasks with hierarchical structure.

What carries the argument

The machinery is a row-wise hyperbolic reparameterization of the frozen weights: HGA maps each row of a layer's Q/V matrix into the Poincaré ball, applies a learnable diagonal Möbius scaling, and maps back, so that a softplus-parameterized vector per layer controls each row's radial displacement. Because of the tangent-radius identity, the effective scaling factor of a row is the ratio of weighted to unweighted Euclidean norms (Proposition 3.2), and because only radial scaling is applied, angular structure is preserved. On the readout side, EMCA gates the pooled audio summary by the prompt, pushes it through three bottleneck branches of different capacities, and fuses them with the Einstein midpoint, whose Lorentz-factor weighting automatically gives more influence to branch points near the ball's boundary; two auxiliary losses, a geodesic prototype loss and a radius-ordering margin loss, anchor that radial structure to emotion classes.

What would settle it

Measure the hyperbolic radius of Qwen2-Audio frame representations against per-segment emotion-label agreement on a benchmark like MELD. If clearly emotional, fine-grained segments sit at smaller radii than neutral or acoustically noisy segments, or if the radius ordering among the three EMCA branches does not correlate with downstream class-separation gains, then the radius-as-granularity assumption is falsified and the reported gains would need another explanation.

Watch

Extended reading notes

Core claim

Adapting the Q/V projection weights of a frozen LALM audio encoder by Möbius scaling in the Poincaré ball, and fusing multi-scale frame summaries with an Einstein midpoint, makes a 0.12%-parameter adapter competitive with or better than Euclidean PEFT. Concretely, HyPASE reports MELD F1 of 53.32% versus 47.84% for LoRA, and IEMOCAP unweighted accuracy of 82.13% versus 78.90% for LoRA, at the cost of a 1.52 pp drop in IEMOCAP weighted accuracy that the authors attribute to hyperbolic space preferentially expanding volume near the boundary, where minority-class representations live. The same adapter transfers zero-shot from MELD to RAVDESS, SAVEE, and IEMOCAP with gains over the frozen Qwen2-Audio baseline.

Load-bearing premise

The load-bearing premise is that the hyperbolic radius of a representation tracks how fine-grained and emotion-specific it is, with coarse acoustic patterns near the ball's center and fine emotion semantics near its boundary, so that radial rescaling of Q/V weight rows and the radius-ordering loss impose the right geometry rather than an arbitrary one.

Editorial extensions

If this is right

  • With roughly 0.12% of parameters trainable, HyPASE improves MELD F1 by 5.48 percentage points over LoRA and beats it on WA and UA as well, suggesting that parameter-efficient speech emotion recognition need not trade away accuracy for budget.
  • On class-imbalanced IEMOCAP, unweighted accuracy rises by 3.23 pp while weighted accuracy falls by 1.52 pp, implying the hyperbolic prior redistributes representational capacity toward underrepresented classes.
  • Zero-shot transfer from MELD to three unseen corpora improves over the frozen Qwen2-Audio baseline, with the largest gain on SAVEE (66.67% versus 30.83% WA).
  • Ablations attribute a 2.09 pp WA gain to the hyperbolic parameterization itself rather than to added structure, and show that the two geometric auxiliary losses are each necessary for the full result.
  • Gromov delta-hyperbolicity measurements of Qwen2-Audio encoder representations, with ratios as low as 0.014 against random Gaussian vectors, support the claim that the encoder's intrinsic geometry is tree-like and therefore suited to hyperbolic adaptation.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the radius-granularity prior generalizes, the same HGA recipe should transfer to other audio-language backbones and to paralinguistic tasks with hierarchical semantics, such as age or speaker-state recognition; a public test would be swapping Qwen2-Audio for another LALM under identical training conditions.
  • The reported IEMOCAP WA/UA trade-off suggests an explicit operating-point knob: curvature or loss-weight schedules could interpolate between majority-favoring and minority-favoring behavior, an axis the paper does not explore.
  • The Gromov delta check is performed on hidden representations, not on the weight rows HGA actually modulates; a direct falsification would measure whether adapted Q/V rows reduce delta or whether radial scaling genuinely separates emotion-class prototypes in hyperbolic distance.
  • Because the text prompt acts only as a task gate, the framework's gains may depend on prompt wording and category ordering; varying the emotion-category list at test time would reveal how much of the transfer is geometric rather than template-specific.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. HyPASE is a parameter-efficient fine-tuning framework for speech emotion recognition in large audio-language models (LALMs). It freezes the Qwen2-Audio-7B backbone and trains two modules: a Hyperbolic Geometric Adapter (HGA) that rescales the frozen query/value weight rows via exponential/logarithmic maps in the Poincaré ball (Eq. 5), and an Emotion-aware Multi-capacity Cross-modal Aggregator (EMCA) that produces audio prefix tokens through task-conditioned gating, three capacity branches, and Einstein midpoint fusion (Eq. 9). The objective (Eq. 11) combines cross-entropy with a prototype-based hyperbolic loss and a radius-ordering loss. Experiments on MELD and IEMOCAP, ablations, zero-shot cross-dataset transfer, and hyperparameter grids are reported.

Significance. If the geometric claims were valid, this would be a notable contribution: it shows strong PEFT performance at roughly 0.12% of backbone parameters, provides extensive ablations and hyperparameter sensitivity analyses, and ships code, pretrained weights, and training logs. The paper's main empirical comparisons are on standard external benchmarks, which mitigates circularity. However, the central geometric mechanism of HGA is mathematically identical to Euclidean diagonal rescaling (Corollary A.5), so the claimed 'hyperbolic' inductive bias for HGA is unsupported. The genuinely hyperbolic component is EMCA's Einstein midpoint fusion, which is a separate downstream aggregation stage. The empirical results remain interesting but need to be reframed or augmented with controls that actually isolate geometry.

major comments (4)
  1. [§3.3 and Appendix A, Corollary A.5] Corollary A.5 proves that the full exp–Möbius–log pipeline for HGA satisfies W_HGA,i = s^(ℓ) ⊙ w_i, i.e., the adapted weight is exactly the frozen weight row multiplied element-wise by a learned vector. Since the exponential and logarithmic maps at the origin are exact inverses, HGA does not implement any hyperbolic operation in function space; it is mathematically a Euclidean diagonal rescaling of the Q/V projections. This directly undermines the statements in §1 and §2.3 that HyPASE is 'the first framework to integrate hyperbolic geometry directly into LALM parameter adaptation' and that HGA performs a 'stronger intervention' than prior fusion-only hyperbolic methods. The ablation M1 vs. G2 in Table 2 (+2.09 pp WA) cannot be attributed to geometry if G2 is also a diagonal scaling; if G2 differs architecturally, then the comparison does not isolate the geometric parameterization. The authors should either modify HGA so that its output depends non-trivially on curvature (e.g., using a nonzero base point for the exponential map, or Möbius addition rather than pure scaling) or substantially soften the hyperbolic claims for HGA.
  2. [§4.3, Table 2] The Euclidean counterparts G1 and G3 are trained with L_CE only, while full HyPASE is trained with L_CE + λ_hyp L_hyp + λ_radius L_radius; the text states that the auxiliary losses are not applied 'because they have no meaningful Euclidean analogue.' This confounds geometry with the loss function. Row L0 in Table 2 (full hyperbolic architecture with L_CE only) reaches 63.90 WA, below G3 (Euclidean architecture, 65.34 WA), and only the addition of the two geometric losses brings the full model to 68.97 WA. To support the claim that 'the performance gap originates from hyperbolic geometry and its supervision,' the Euclidean counterparts should be trained with analogous Euclidean losses—e.g., a prototype softmax on Euclidean distances and a radius-ordering loss on branch output norms—under matched training schedules.
  3. [§4.2–4.5, Tables 1–4] No error bars, repeated-seed statistics, or significance tests are reported for any table. Differences such as the +2.09 pp WA for M1 vs. G2 in Table 2 and the +3.23 pp UA gain on IEMOCAP in Table 1 are small relative to the typical variance of PEFT fine-tuning on a frozen 7B backbone. A single run per configuration cannot establish that these differences are reliable. Please report mean and standard deviation over at least three seeds for the main results and the key ablations, or provide statistical significance tests.
  4. [§3.1.1, Eq. (13), and §4.6] The central premise—that hyperbolic radius is a valid proxy for representational granularity, with smaller radii for coarse patterns and larger radii for fine-grained emotion semantics—is asserted in Section 3.1.1 and operationalized in the L_radius ordering loss (Eq. 13), but it is not directly tested. The Gromov δ-hyperbolicity analysis in Section 4.6.1 shows that audio representations have lower δ than random Gaussian vectors, but this does not establish the radius-to-granularity mapping. The layer-wise scaling analysis in Section 4.6.2 shows only that learned scaling factors vary by layer, not that radius tracks emotionally meaningful granularity. A concrete test is needed, such as correlating learned branch or prototype radii with emotion-class granularity, or demonstrating that coarse-versus-fine emotion cues occupy systematically different radii in the Poincaré ball.
minor comments (5)
  1. [Appendix A] Appendix A refers to 'Theorem 1 and Proposition 1' but the main text numbers these as Theorem 3.1 and Proposition 3.2; the numbering should be consistent.
  2. [§4.6.2] The sentence saying that deeper layers have larger scaling factors is 'consistent with the geometric prediction of Proposition 3.2' is imprecise: Proposition 3.2 describes row-wise effective radius scaling, not a layer-wise ordering; please rephrase.
  3. [§2.3 and References] The text 'HyFuse [13, 42]' cites two references, but [13] is 'Emotion Collider' and only [42] is 'HyFuse'; please correct the citation grouping.
  4. [Appendix E] The Gromov δ-hyperbolicity computation is not described: the estimator, the number of sampled points, and the graph construction used to compute δ are all unspecified; without these details the reported δ_audio/δ_random ratios cannot be reproduced.
  5. [§4.4] The claim that HyPASE 'remains comparatively stable' across zero-shot corpora is based on a single run per corpus; stability cannot be assessed without variance information.

Circularity Check

1 steps flagged · score 6.0 of 10

HGA's exp/log pipeline collapses to element-wise scaling (Corollary A.5), so the core 'hyperbolic' adapter is functionally a Euclidean diagonal rescaling; the geometric attribution reduces by construction, though EMCA retains genuine hyperbolic fusion.

  1. renaming known result [Appendix A, Corollary A.5 (see also Eq. (5) and the implementation paragraph in Section 3.3.1)]
    "Corollary A.5 (HGA forward-pass simplification). The HGA Euclidean weight matrix satisfies W^{(ℓ)}_{HGA,i} = log^{D,c}_0(w^{(ℓ)}_{D,i}) = s^{(ℓ)} ⊙ w_i, i.e., the full exp–Möbius–log pipeline reduces to element-wise scaling of the frozen weight rows in tangent space."

    Equation (5) defines HGA as log( W_s ⊗_c exp(w_0,i) ). Because diagonal Möbius multiplication is exp(s ⊙ log(·)) and log is exp's exact inverse, the composition collapses to s⊙w_i, exactly as Corollary A.5 states. HGA's 'hyperbolic radius modulation' is therefore a Euclidean diagonal rescaling of the frozen Q/V rows by construction; the exp/log maps cancel. Yet the paper calls HGA the 'core geometric contribution' and uses the M1-vs-G2 ablation (+2.09 pp WA) to attribute a benefit to 'hyperbolic' vs 'Euclidean' HGA. Since the two operations are the same function class, any gain cannot be due to hyperbolic geometry. The claimed HGA hyperbolic inductive bias reduces by definition to element-wise scaling.

full rationale

The empirical evaluation is anchored to external benchmarks (MELD, IEMOCAP) and an external backbone (Qwen2-Audio), and the performance numbers are not predicted from the geometric theory; they are fitted and measured. There is no load-bearing self-citation chain: prior hyperbolic works cited (HyperET, HySAC, HGCN) are by other authors, and the Gromov delta measurements are presented as external evidence rather than assumed. The circularity burden is concentrated in HGA. The paper's own Corollary A.5 proves that the exp–Möbius–log pipeline of Eq. (5) is exactly element-wise scaling of the frozen weight rows in tangent space. Since the Poincaré exponential/logarithm at the origin are inverse diffeomorphisms, the hyperbolic radius modulation of HGA has no functional effect beyond a diagonal rescaling available in Euclidean space. The M1-vs-G2 ablation therefore cannot isolate a geometric inductive bias; the +2.09 pp difference must stem from parameterization, initialization, or numerics, not from hyperbolic geometry. EMCA's Einstein midpoint and the Lhyp/Lradius losses do involve genuine Poincaré geometry, so the paper is not wholly vacuous; these components are separate and do not rescue the HGA claim. The result is a partial circularity: the central geometric mechanism for the HGA half of the framework reduces by construction to a Euclidean operation, while the overall empirical claims remain independent.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

HyPASE introduces no new particle-like entities, forces, or conserved quantities. The entities it introduces are architectural: a hyperbolic geometric adapter in weight space, a task-conditioned gate, three capacity branches, and a set of class prototypes. The main conceptual assumption, the radius-as-granularity proxy, is an interpretation imposed on a geometric quantity and is the load-bearing part of the framework. The free parameters are standard hyperparameters, but the sensitivity appendix shows that several of them sit near cliff edges, so they are not innocent choices.

free parameters (6)
  • Curvature c = 0.01
    Default curvature chosen without theory; sensitivity sweep in Appendix D shows 0.03 is workable and 0.1 collapses to near-random performance on the EMCA-only branch.
  • Auxiliary loss weight lambda_hyp = 0.1
    Selected from the lambda_hyp x lambda_radius grid in Table 8; the 0.05 value can be harmful when paired with lambda_radius 0.005.
  • Auxiliary loss weight lambda_radius = 0.01
    Grid-selected; the sensitivity sweep shows that 0.02 collapses branch ordering and degrades WA to near-random levels.
  • Temperature tau = 0.07
    Main experiments use 0.07 while the appendix grid uses 0.1; the sensitivity table shows strong cliff edges at 0.2 and above, so the exact value matters.
  • Radius ranking margin m = 0.5
    Hyperparameter of the Lradius ranking loss; chosen by hand with no sensitivity analysis reported.
  • Architecture capacities = r=12, d_h=256, K=4, branch ratios 1/8, 1/16, 1/32
    These are design choices that determine the parameter count and the radial ordering scaffold; no ablation on r, d_h, or K is reported.
assumptions (5)
  • standard math The Poincare ball exponential/log maps, Mobius scalar multiplication, and Einstein midpoint correctly implement hyperbolic operations.
    Used throughout Sections 3.1 and 3.4, with a self-contained operation reference in Appendix B. These are standard results from the cited hyperbolic learning literature.
  • ad hoc to paper Hyperbolic radius is a valid proxy for representational granularity, with smaller radii for coarse patterns and larger radii for fine-grained emotion semantics.
    This is the core design assumption introduced in Section 3.1.1. The paper motivates it with the Gromov delta analysis, but the analysis does not directly establish that radius tracks emotionally meaningful granularity.
  • domain assumption Qwen2-Audio encoder representations evolve monotonically from coarse acoustic templates to fine emotion-discriminative abstractions across depth.
    Stated in Section 1 and used to justify layer-adaptive HGA scaling. Figure 4 shows lower Gromov delta than random Gaussians, but not a semantic coarse-to-fine ordering.
  • ad hoc to paper Freezing K while modulating Q/V preserves the pre-trained attention structure and only shifts representational granularity.
    Argued in Section 3.3.2 as a deliberate asymmetric design. The paper provides no direct comparison against a version that also modulates K.
  • ad hoc to paper Higher-capacity EMCA branches should occupy larger hyperbolic radii than lower-capacity branches.
    This is a direct consequence of the Lradius loss in Equation 13, which enforces monotonic branch ordering. The later observation of radial ordering is therefore partially enforced rather than emergent.

how reviews work

0 comments
Cite this review

Pith. "Pith review of HyPASE: Hyperbolic Geometry for Parameter-Efficient Speech Emotion Fine-Tuning Framework for Large Audio-Language Models." pith.science (2026). https://pith.science/paper/XE6ZBCEG

@misc{pith2026260804351,
  author       = {Pith},
  title        = {Pith review of: HyPASE: Hyperbolic Geometry for Parameter-Efficient Speech Emotion Fine-Tuning Framework for Large Audio-Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XE6ZBCEG}},
  note         = {Machine review of arXiv:2608.04351}
}
read the original abstract

Large Audio-Language Models (LALMs) excel at general speech understanding; however, adapting them to fine-grained tasks like Speech Emotion Recognition (SER) remains a significant bottleneck. Current Parameter-Efficient Fine-Tuning (PEFT) methods typically operate in flat Euclidean space, and this geometry fails to capture the multi-granularity nature of emotion cues, which range from low-level prosody to high-level semantics. To address this, we propose HyPASE, a hyperbolic PEFT framework for LALM-based SER. HyPASE leverages the Poincare ball model, using the hyperbolic radius as an explicit proxy for representational granularity. The framework consists of two core components: a Hyperbolic Geometric Adapter (HGA) for layer-adaptive weight modulation, and an Emotion-aware Multi-capacity Cross-modal Aggregator (EMCA) that compresses multi-scale features into compact audio prefixes. Empirical results on standard benchmarks show that HyPASE outperforms Euclidean PEFT baselines across all metrics on MELD and achieves a notable Unweighted Accuracy gain on IEMOCAP, particularly in class-imbalanced emotion recognition, with the accompanying slight Weighted Accuracy trade-off reflecting hyperbolic space's geometric prioritization of minority-class representations; furthermore, HyPASE achieves robust zero-shot cross-dataset generalization within a constrained parameter budget. By grounding the adaptation process in hyperbolic geometry, HyPASE offers a highly efficient path for LALM fine-tuning.

Figures

Figures reproduced from arXiv: 2608.04351 by the authors.

Figure 1
Figure 1. Overall framework of HyPASE. HGA is injected into each layer of the frozen audio encoder, mapping Q/V weight [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Method details of HyPASE. Left: HGA maps frozen [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Distribution of HGA diagonal scaling parameters [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: reports the full per-layer Gromov 𝛿-hyperbolicity com￾parison referenced in the main paper (Section 4.5): the true audio representation 𝛿audio against the random Gaussian baseline 𝛿random at every sampled layer of the Qwen2-Audio audio encoder. 0 3 7 11 15 19 23 27 31 …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 27 canonical work pages

  1. [1]

    Alexei Baevski, Wei-Ning Hsu, Alexis Conneau, and Michael Auli. 2023. data2vec 2.0: Highly Efficient Self-Supervised Learning for Vision, Speech, and Text.arXiv preprint arXiv:2212.07525(2023)

    arXiv 2023

  2. [2]

    Rainer Banse and Klaus R. Scherer. 1996. Acoustic profiles in vocal emotion expression.Journal of Personality and Social Psychology70, 3 (1996), 614–636. doi:10.1037/0022-3514.70.3.614 10 HyPASE: Hyperbolic Geometry for Parameter-Efficient Speech Emotion Fine-Tuning Framework for Large Audio-Language Models

    work page doi:10.1037/0022-3514.70.3.614 1996

  3. [3]

    Chang, Sungbok Lee, and Shrikanth S

    Carlos Busso, Murtaza Bulut, Chi-Chun Lee, Abe Kazemzadeh, Emily Mower, Samuel Kim, Jeannette N. Chang, Sungbok Lee, and Shrikanth S. Narayanan. 2008. IEMOCAP: interactive emotional dyadic motion capture database.Language Resources and Evaluation42, 4 (Nov. 2008), 335–359. doi:10.1007/s10579-008- 9076-6

    work page doi:10.1007/s10579-008- 2008

  4. [4]

    Haoxiang Cao, Chaoqun Wang, Yongwen Lai, Shaobo Min, and Xuejin Chen

    work page

  5. [5]

    Ines Chami, Zhitao Ying, Christopher Ré, and Jure Leskovec. 2019. Hyperbolic Graph Convolutional Neural Networks. InAdvances in Neural Information Pro- cessing Systems (NeurIPS). 4868–4879. https://proceedings.neurips.cc/paper/ 2019/hash/0415740eaa4d9decbc8da001d3fd805f-Abstract.html

    arXiv 2019

  6. [6]

    Sanyuan Chen, Chengyi Wang, Zhengyang Chen, Yu Wu, Shujie Liu, Zhuo Chen, Jinyu Li, Naoyuki Kanda, Takuya Yoshioka, Xiong Xiao, et al . 2022. WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing.IEEE Journal of Selected Topics in Signal Processing16, 6 (2022), 1505–1518

    work page 2022

  7. [7]

    Yunfei Chu, Jin Xu, Qian Zhou, Zhifu Yang, Zhicheng Shao, Zheng Yang, Jingren Lin, and Chang Zhou. 2024. Qwen2-Audio Technical Report.arXiv preprint arXiv:2407.10759(2024). https://arxiv.org/abs/2407.10759

    arXiv 2024

  8. [8]

    Kuiye Ding, Fanda Fan, Yao Wang, Ruijie Jian, Xiaorui Wang, Luqi Gong, Yishan Jiang, Chunjie Luo, and Jianfeng Zhan. 2025. DualSG: A Dual-Stream Explicit Semantic-Guided Multivariate Time Series Forecasting Framework. InProceed- ings of the 33rd ACM International Conference on Multimedia. 508–517

    work page 2025

Show all 45 references
  1. [9]

    Hongfei Du, Sidi Lu, Gang Zhou, and Ye Gao. 2025. EAA: Emotion-Aware Audio Large Language Models with Dual Cross-Attention and Context-Aware Instruction Tuning. InInterspeech 2025. 5433–5437. doi:10.21437/Interspeech. 2025-1232

    2025 doi

  2. [10]

    Soumya Dutta and Sriram Ganapathy. 2024. Leveraging Content and Acoustic Representations for Speech Emotion Recognition. doi:10.48550/ARXIV.2409. 05566

    2024 doi

  3. [11]

    Scherer, Björn W

    Florian Eyben, Klaus R. Scherer, Björn W. Schuller, Johan Sundberg, Elisabeth André, Carlos Busso, Laurence Y. Devillers, Julien Epps, Petri Laukka, Shrikanth S. Narayanan, and Khiet P. Truong. 2016. The Geneva Minimalistic Acoustic Parameter Set (GeMAPS) for Voice Research an...

    2016

  4. [12]

    Lihuang Fang, Yuchen Zou, Kebing Jin, and Jinghui Qin. 2026. EmoAgent-R1: Towards Multimodal Emotion Understanding with Reinforcement Learning- based Dynamic Agent Specialization. InProceedings of the 34th ACM International Conference on Multimedia

    2026

  5. [13]

    Rong Fu, Ziming Wang, Shuo Yin, Kun Liu, Xianda Li, and Simon Fong. 2026. Emotion Collider: Dual Hyperbolic Mirror Manifolds for Sentiment Recovery via Anti Emotion Reflection. arXiv:2602.16161 [cs.MM] https://arxiv.org/abs/ 2602.16161

    2026 arXiv

  6. [14]

    Octavian-Eugen Ganea, Gary Bécigneul, and Thomas Hofmann. 2018. Hyperbolic Neural Networks. InAdvances in Neural Information Processing Systems (NeurIPS)

    2018

  7. [15]

    Christer Gobl and Ailbhe Ní Chasaide. 2003. The role of voice quality in com- municating emotion, mood and attitude.Speech Communication40, 1–2 (April 2003), 189–212. doi:10.1016/S0167-6393(02)00082-1

    2003 doi

  8. [16]

    Sanaul Haq and Philip J.B. Jackson. 2011. Visual Speech Recognition Using Optical Flow and Hidden Markov Models. InAutonomous Agents and Multi- Agent Systems. SAVEE database

    2011

  9. [17]

    Neil Houlsby, Andrei Giurgiu, Stanislaw Jastrzebski, Bruna Morrone, Quentin De Laroussilhe, Andrea Gesmundo, Mona Attariyan, and Sylvain Gelly. 2019. Parameter-Efficient Transfer Learning for NLP. InInternational Conference on Ma- chine Learning (ICML). 2790–2799. http://proce...

    2019

  10. [18]

    Wei-Ning Hsu, Benjamin Bolte, Yao-Hung Hubert Tsai, Kushal Lakhotia, Ruslan Salakhutdinov, and Abdelrahman Mohamed. 2021. HuBERT: Self- Supervised Speech Representation Learning by Masked Prediction of Hidden Units.IEEE/ACM Transactions on Audio, Speech, and Language Processin...

    2021

  11. [19]

    Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen

    Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2022. LoRA: Low-Rank Adaptation of Large Language Models. InInternational Conference on Learning Representations (ICLR). https://openreview.net/forum?id=nZeVKeeFYf9

    2022

  12. [20]

    Pauline Larrouy-Maestri, David Poeppel, and Marc D. Pell. 2024. The Sound of Emotional Prosody: Nearly 3 Decades of Research and Future Directions. Perspectives on Psychological Science20, 4 (Jan. 2024), 623–638. doi:10.1177/ 17456916231217722

    2024

  13. [21]

    Pengcheng Li, Botao Zhao, Zuheng Kang, Junqing Peng, Xiaoyang Qu, Yayun He, and Jianzong Wang. 2025. EMO-RL: Emotion-Rule-Based Reinforcement Learning Enhanced Audio-Language Model for Generalized Speech Emotion Recognition. InFindings of the Association for Computational Ling...

    2025

  14. [22]

    Xiang Lisa Li and Percy Liang. 2021. Prefix-Tuning: Optimizing Continuous Prompts for Generation. InAnnual Meeting of the Association for Computational Linguistics (ACL). 4582–4597. doi:10.18653/v1/2021.acl-long.353

    2021 doi

  15. [23]

    doi:10.18653/v1/2025.findings-emnlp.1018

    Association for Computational Linguistics, Suzhou, China, 18744–18754. doi:10.18653/v1/2025.findings-emnlp.1018

    2025 doi

  16. [24]

    Livingstone and Frank A

    Steven R. Livingstone and Frank A. Russo. 2018. The Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS): A Dynamic, Multimodal Set of Facial and Vocal Expressions in North American English.PLoS ONE13, 5 (2018), e0196391

    2018

  17. [25]

    Zhipeng Li, Xiaofen Xing, Yuanbo Fang, Weibin Zhang, Hengsheng Fan, and Xiangmin Xu. 2024. Multi-Scale Temporal Transformer For Speech Emotion Recognition.arXiv preprint arXiv:2410.00390(2024). https://arxiv.org/abs/2410. 00390

    2024 arXiv

  18. [26]

    Ziyang Ma, Zhisheng Zheng, Jiaxin Ye, Jinchao Li, Zhifu Gao, Shiliang Zhang, and Xie Chen. 2024. emotion2vec: Self-Supervised Pre-Training for Speech Emotion Representation. InACL Findings

    2024

  19. [27]

    Ziyang Ma, Zhisheng Zheng, Jiaxin Ye, Jinchao Li, Zhifu Gao, Shiliang Zhang, and Xie Chen. 2024. EmoBox: Multilingual Multi-corpus Speech Emotion Recognition Toolkit and Benchmark.arXiv preprint arXiv:2406.07162(2024)

    2024 arXiv

  20. [28]

    Maximilian Nickel and Douwe Kiela. 2017. Poincaré Embeddings for Learning Hierarchical Representations. InAdvances in Neural Information Processing Sys- tems (NeurIPS). 6338–6347. https://proceedings.neurips.cc/paper/2017/hash/ 59dfa2df42d9e3d41f5b02bfc32229dd-Abstract.html

    2017

  21. [29]

    Jialong Mai, Xiaofen Xing, Yawei Li, Weidong Chen, Zhipeng Li, Jingyuan Xing, and Xiangmin Xu. 2025. Dynamic Parameter Memory: Temporary LoRA- Enhanced LLM for Long-Sequence Emotion Recognition in Conversation.arXiv preprint arXiv:2507.09076(2025). https://arxiv.org/abs/2507.09076

    2025

  22. [30]

    Tobia Poppi, Tejaswi Kasarla, Pascal Mettes, Lorenzo Baraldi, and Rita Cuc- chiara. 2025. Hyperbolic Safety-Aware Vision-Language Models.arXiv preprint arXiv:2503.12127(2025). https://arxiv.org/abs/2503.12127 CVPR 2025

    2025 arXiv

  23. [31]

    Zelin Peng, Zhengqin Xu, Qingyang Liu, Xiaokang Yang, and Wei Shen. 2025. HyperET: Efficient Training in Hyperbolic Space for Multi-modal Large Language Models.arXiv preprint arXiv:2510.20322(2025). https://arxiv.org/abs/2510.20322 NeurIPS 2025 Oral

    2025

  24. [32]

    Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, and Ilya Sutskever. 2023. Robust Speech Recognition via Large-Scale Weak Supervision.arXiv preprint arXiv:2212.04356(2023). Whisper

    2023 arXiv

  25. [33]

    Soujanya Poria, Devamanyu Hazarika, Navonil Majumder, Gautam Naik, Erik Cambria, and Rada Mihalcea. 2019. MELD: A Multimodal Multi-Party Dataset for Emotion Recognition in Conversations. InAnnual Meeting of the Association for Computational Linguistics (ACL). 527–536. doi:10.1...

    2019 doi

  26. [34]

    Changli Tang, Wenyi Yu, Guangzhi Sun, Xianzhao Chen, Tian Tan, Wei Li, Lu Lu, Zejun Ma, and Chao Zhang. 2023. SALMONN: Towards Generic Hearing Abilities for Large Language Models.arXiv preprint arXiv:2310.13289(2023). https://arxiv.org/abs/2310.13289

    2023 arXiv

  27. [35]

    Klaus R. Scherer. 1986. Vocal affect expression: A review and a model for future research.Psychological Bulletin99, 2 (1986), 143–165. doi:10.1037/0033-2909.99.2. 143

    1986 doi

  28. [36]

    Schuller

    Panagiotis Tzirakis, Anh Nguyen, Stefanos Zafeiriou, and Bjorn W. Schuller. 2021. Speech Emotion Recognition Using Semantic Information. InICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). 6279–6283. doi:10.1109/ICASSP39728.2...

    2021

  29. [37]

    Hugo Thimonier, Antony Perzo, and Renaud Seguier. 2025. EmoSLLM: Parameter- Efficient Adaptation of LLMs for Speech Emotion Recognition.arXiv preprint arXiv:2508.14130(2025). https://arxiv.org/abs/2508.14130

    2025 arXiv

  30. [38]

    Junhao Xiao, Shun Feng, Zhiyu Wu, Jinghan Yu, Haibiao Yao, Zhiyuan Ma, Jianjun Li, Youjun Bao, and Yi Chen. 2026. Dual-Stream Decoupled Learning for Temporal Consistency and Speaker Interaction in AVSD. arXiv:2512.19130 [cs.MM] https: //arxiv.org/abs/2512.19130

    2026 arXiv

  31. [39]

    Zhaohui Wang, Tengbo Yu, and Hao Tang. 2025. CoT4AD: A Vision-Language- Action Model with Explicit Chain-of-Thought Reasoning for Autonomous Driv- ing.arXiv preprint arXiv:2511.22532(2025)

    2025

  32. [40]

    Yiqing Yang and Man-Wai Mak. 2025. EmoQ: Speech Emotion Recogni- tion via Speech-Aware Q-Former and Large Language Model.arXiv preprint arXiv:2509.15775(2025). https://arxiv.org/abs/2509.15775

    2025

  33. [41]

    Ho, and Hung-yi Lee

    Chih-Kai Yang, Neo S. Ho, and Hung-yi Lee. 2025. Towards Holistic Evaluation of Large Audio-Language Models: A Comprehensive Survey. InProceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics, Suzhou, Chi...

    2025 doi

  34. [42]

    Yixuan Zhang et al. 2025. HyFuse: Hyperbolic Multimodal Fusion for Speech Emotion Recognition.arXiv preprint arXiv:2506.03403(2025)

    2025 arXiv

  35. [43]

    Ziyou Yang, Yu Wu, Xinyu Jiang, Zhenghan Wu, Xiaoyu Yang, Yuxin Zhao, Yue Wang, Chaoran Cui, Zhen Lian, Bin Liu, et al. 2025. MSE-Adapter: A Lightweight Plugin Endowing LLMs with the Capability to Perform Multimodal Sentiment Analysis and Emotion Recognition.arXiv preprint arX...

    2025 arXiv

  36. [45]

    Zhixian Zhao, Xinfa Zhu, Xinsheng Wang, Shuiyuan Wang, Xuelong Geng, Wenjie Tian, and Lei Xie. 2025. Steering Language Model to Stable Speech Emotion Recognition via Contextual Perception and Chain of Thought.arXiv preprint arXiv:2502.18186(2025). doi:10.1109/TASLPRO.2025.3648793 11

    2025

  37. [2025]

    InProceedings of the 33rd ACM International Conference on Multimedia

    Causalctrl: Causality-aware control framework for text-guided visual editing. InProceedings of the 33rd ACM International Conference on Multimedia. 9920–9929

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.