Pith. sign in

REVIEW 3 major objections 5 minor 59 references

ODE-Based Transformer Decoders for Iterative Sign Language Translation

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read The paper claims that replacing each residual refinement step in an iterative sign-language decoder with a Runge-Kutta update (RK-2 or RK-4) improves translation quality on PHOENIX-14T and CSL-Daily without adding decoder parameters.

desk verdict Honest, well-reported RK/ODE refinement for sign-language translation with clean ablations, but the headline gains are confounded with extra function evaluations and unquantified variance. read the letter →

arxiv 2608.11352 v1 pith:C7GVKPAW submitted 2026-08-11 cs.CL cs.LG

Tuğçe K{i}z{i}ltepe , Hacer Yalim Keles This is my paper · ORCID
classification cs.CLcs.LG
keywords signlanguagetranslationiterativerefinementRunge-KuttamethodsneuralODETransformerdecoderparameter-efficientgloss-freeBLEU
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

Sign language translation systems have improved mainly by scaling up pretrained encoders and language models. This paper proposes a parameter-free way to get more out of the same-sized decoder: treat each residual update in an iterative refinement decoder as one Euler step of a continuous dynamical system, and replace it with a second- or fourth-order Runge-Kutta step. On the PHOENIX-14T and CSL-Daily benchmarks this beats the reproduced IPSLT iterative refinement baseline in BLEU-4, and on CSL-Daily it does so with fewer decoder layers and refinement iterations. The significance, if the claim holds, is that translation quality can improve by changing how representations are updated rather than by adding capacity.

What carries the argument

The load-bearing mechanism is the reinterpretation of residual decoder updates as discretized ODE flow, combined with classical Runge-Kutta coefficients. In the direct RK-2 formulation, one decoder layer $F$ is evaluated twice per update, $F_1=F(y_t)$ and $F_2=F(y_t+F_1)$, and the step is $y_{t+1}=y_t+\tfrac12(F_1+F_2)$; the residual-compatible formulation recovers $F(y)=g(y)-y$ from a standard Transformer layer $g$, and RK-4 performs four stage evaluations with weights $\tfrac16,\tfrac13,\tfrac13,\tfrac16$. All stage evaluations reuse the same decoder function, so the order of the scheme rises without adding parameters. A post-update LayerNorm and an iteration-wise distillation loss are optional stabilizers; the ablations show their best setting is dataset-dependent.

What would settle it

Train a matched control decoder that performs the same number of function evaluations per refinement step as RK-2 or RK-4 but with plain repeated Euler/residual updates (or shuffled stage coefficients), holding parameters, FLOPs, and training schedule fixed. If BLEU-4 does not fall, the reported gains come from extra computation, not from ODE-accurate integration.

Watch

Extended reading notes

Core claim

The central discovery claim is that higher-order numerical integration of the decoder's latent update dynamics improves sign language translation quality without adding decoder parameters. The paper frames the standard residual update $y_{t+1}=y_t+F(y_t)$ as a first-order Euler discretization of an underlying ODE, then applies RK-2 and RK-4 schemes, either directly on the decoder layer $F$ (Method 1) or through the residual behavior of a standard Transformer layer $g(y)=y+F(y)$ (Method 2). With the same visual backbone and training setup as the reproduced IPSLT baseline, RK2SLT-M1 reaches 22.96 BLEU-4 on PHOENIX-14T and RK2SLT-M2 reaches 19.34 BLEU-4 on CSL-Daily, topping IPSLT on both test sets. The paper also reports that the RK-4 variant RK4SLT-M2 comes within 0.09 BLEU-4 of the best CSL-Daily result while using a single decoder layer and a single refinement iteration, which it reads as evidence that better per-step dynamics can substitute for depth and repeated refinement.

Load-bearing premise

The load-bearing premise is that the residual updates of the decoder are well approximated by a smooth continuous dynamical system, so that Runge-Kutta coefficients improve the accuracy of the representation update rather than merely adding function evaluations.

Editorial extensions

If this is right

  • With the same visual backbone and training protocol, RK-2 and RK-4 updates beat the reproduced IPSLT baseline on PHOENIX-14T and CSL-Daily, so the reported gains are tied to update dynamics rather than model capacity.
  • Higher-order updates create accuracy-efficiency operating points: on CSL-Daily, RK4SLT-M2 stays within 0.09 BLEU-4 of the best variant while using one decoder layer and one refinement iteration, trading per-step compute for reduced depth.
  • The parameter-efficient claim holds with active inference parameters actually reduced: RK2SLT-M2 cuts inference-path parameters by 12.6% on PHOENIX-14T and 7.4% on CSL-Daily relative to IPSLT.
  • The best ODE-decoder configuration is dataset-dependent, so deployment must tune placement, LayerNorm, distillation weight, iteration count, and decoder depth for the target data.
  • The approach offers a complementary direction to scaling: future quality gains in sign language translation can come from stronger refinement dynamics rather than larger pretrained backbones.

Reading between the lines

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

  • If the smooth-flow interpretation is right, adaptive step-size or order selection at inference (choosing the refinement count or RK order per sequence) should improve the accuracy-efficiency frontier further; the paper does not test this.
  • The decoder update is task-agnostic, so the same RK swap could transfer to other iterative refinement sequence tasks such as non-autoregressive machine translation or image captioning.
  • The finding that RK-4's extra evaluations substitute for decoder depth suggests a testable compute-equivalence hypothesis: one fourth-order step may be worth several repeated first-order steps at matched FLOPs.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes replacing the standard residual decoder update in an iterative sign language translation decoder with Runge-Kutta-inspired updates (RK-2 and RK-4), framed as higher-order numerical integration of an underlying ODE. The authors report BLEU-4 gains over a reproduced IPSLT baseline on PHOENIX-14T and CSL-Daily without adding decoder parameters, and they provide parameter counts, FLOPs, and latency figures to separate parameter efficiency from computational efficiency. Configurations are selected through sequential DEV-set ablations, and the final test results are reported for the selected variants.

Significance. The contribution is potentially useful but modest in scope: if the reported gains are actually due to the higher-order update rule rather than to the extra decoder function evaluations, the work demonstrates a parameter-efficient complement to scaling on two standard SLT benchmarks. The paper deserves credit for reproducing the IPSLT baseline under a matched training/inference setup, reporting transparent parameter/FLOP/latency accounting, running detailed sequential ablations, and honestly distinguishing parameter efficiency from computational efficiency in the conclusion. The main empirical claim, however, needs a matched-compute control and statistical validation before the ODE-mechanism explanation can be considered supported.

major comments (3)
  1. [Section 3.4, Tables 1-4] The central comparison does not isolate the effect of the RK update rule from the increased number of decoder function evaluations. Because RK-2 and RK-4 evaluate the shared decoder function fθ two and four times per refinement step, while the IPSLT baseline performs one evaluation per step, Table 2 shows RK2SLT-M1 uses 81.97 G FLOPs versus 58.25 G for IPSLT on PHOENIX-14T. No Euler baseline with a matched FLOP budget (for example, repeated Euler substeps with the same number of fθ evaluations per decoder step) is reported, so the BLEU gains (22.96 vs 21.82 on PHOENIX-14T) could be attributed to additional shared-weight forward passes rather than to higher-order integration accuracy. Please add a matched-compute Euler control and, if the gains persist, report it as evidence for the ODE-mechanism claim.
  2. [Section 3.4, Method 1 equations] The direct RK-2 formulation as written is mathematically ambiguous. If F denotes the full decoder layer g with residual connections, so that g(y)=y+F_residual(y), then the stated update y_{t+1}=y_t+1/2(F1+F2) with F1=g(y_t) and F2=g(y_t+F1) does not reduce to an RK-2 update of the residual increment and appears to double-count the identity. If F is intended to denote the residual function only, the text should say so explicitly and specify how it is extracted from a standard decoder layer, since the layer itself already contains residuals. The same ambiguity carries over to the RK-4 direct formulation in Eq. (3). Please clarify the exact forward computation for Method 1 and align the notation with Eq. (2), or provide pseudocode for the implemented variant.
  3. [Section 4.2, Tables 5-8 and Tables 1/3] No repeated runs, standard deviations, or significance tests are reported, and the best configuration is selected through a sequential greedy search on the DEV set. The headline improvements over the reproduced IPSLT baseline are small (0.64-1.14 BLEU-4 on PHOENIX-14T and 0.47-0.86 BLEU-4 on CSL-Daily), so run-to-run variance and selection bias could account for part of the reported advantage. Please report mean and standard deviation over at least three random seeds for the main comparisons and apply a paired significance test (e.g., sentence-level bootstrap resampling) to the headline BLEU-4 differences.
minor comments (5)
  1. [Section 3.4] The symbol F is used both for the residual increment in Eq. (2) and for the full decoder-layer transformation in Method 1; this notation conflict should be resolved globally to make the derivation reproducible.
  2. [Table 5] The learnable RK-2 combination coefficient α is a new free parameter; the 'without adding parameters' phrasing should state explicitly that this is a scalar parameter outside the decoder rather than claiming no parameters at all are added.
  3. [Section 4.1] No code, checkpoints, or random seeds are provided; please include code or a detailed configuration listing to support reproducibility of the reported BLEU-4 numbers.
  4. [Section 2.2] Since RK-inspired Transformer updates already appear in the cited work by Li et al. [25], the 'first application' claim should be phrased as the first application to iterative sign language translation refinement, not as introducing the ODE/RK viewpoint to sequence generation.
  5. [Tables 1 and 3] The SOTA rows use approximate parameter counts and different pretraining, visual encoders, and evaluation conditions; the joint comparison with those rows should be framed only as context, while the matched IPSLT reproduction is the appropriate basis for the parameter-efficiency argument.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the RK-2/RK-4 updates are standard numerical schemes applied to an explicitly written residual update, and the reported BLEU scores are empirical outcomes, not derived from the RK coefficients.

full rationale

The paper's claimed derivation chain is not circular. In Section 3.4 the residual update y_{t+1}=y_t+F(y_t) is written explicitly and reinterpreted as a first-order Euler step; the RK-2 and RK-4 formulas then use the classical, externally defined Runge-Kutta coefficients (1/2,1/2 and 1/6,1/3,1/3,1/6) applied to repeated evaluations of the same decoder function. These coefficients are not fitted to any data and are not derived from the BLEU-4 numbers in Tables 1 and 3; the test scores are empirical outcomes. The ODE interpretation is explicitly inherited from independent prior work [25,58], and the paper states it is 'Building on Li et al. [25]' rather than importing a uniqueness theorem or using a self-citation to justify the method. The ablations select hyperparameters by DEV BLEU-4 and then report held-out TEST BLEU-4, which is standard model selection rather than a fitted parameter renamed as a prediction. The authors' own prior work is cited only for related iterative refinement in sign language production [24] and an earlier dataset [43], not as load-bearing support for the RK update. The main residual risk is experimental: RK-2/RK-4 perform 2-4 decoder evaluations per step while IPSLT performs one, so the reported gains are not isolated from added function evaluations; but that confound is a question of internal validity and control conditions, not a reduction of the prediction to its inputs by construction.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The paper makes no theoretical derivation; its causal claim rests on the ODE analogy and on the empirical success of iterative refinement. All tunable design choices are selected on dev data, and the learnable RK-2 coefficient is trained rather than derived.

free parameters (4)
  • RK-2 learnable combination coefficient alpha = not reported (selected on DEV)
    Used in alpha*F1 + (1-alpha)*F2 variant; learned during training and chosen via dev BLEU; affects one of the four reported variants, not the core averaged formulation.
  • Iterative distillation loss weight (IDL) = 0.05, 0.1, 0.2, or 0.3 depending on dataset and variant
    Chosen by dev-set grid; values reported in Tables 5-8.
  • Refinement iteration count K = 1 to 3 per variant and dataset
    Selected by dev BLEU-4; central to the efficiency claims.
  • Decoder depth = 1 to 3 layers per variant and dataset
    Selected by dev BLEU-4; drives parameter and FLOP differences.
assumptions (4)
  • standard math Classical RK-2 and RK-4 coefficient sets are valid numerical integration schemes for ODEs.
    Invoked in Eq. 3 and surrounding definitions in Section 3.4; no derivation of coefficients is provided.
  • domain assumption Residual Transformer update y_{t+1}=y_t+F(y_t) is an Euler discretization of an underlying continuous ODE.
    This is the central analogy in Section 3.4; if it fails, the higher-order integration interpretation loses its justification.
  • domain assumption The decoder layer F can be evaluated multiple times per step as a shared function, so RK intermediate stages are valid updates of the same dynamics.
    Section 3.4 uses repeated applications of the same f_theta to compute F1 through F4; shared-weight multi-evaluation is an algorithmic choice that may not match a true continuous flow.
  • domain assumption Iterative refinement with K steps improves translation quality over a single pass.
    Borrowed from IPSLT [52], Section 2.3; the entire comparison is against the IPSLT iterative refinement baseline.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ODE-Based Transformer Decoders for Iterative Sign Language Translation." pith.science (2026). https://pith.science/paper/C7GVKPAW

@misc{pith2026260811352,
  author       = {Pith},
  title        = {Pith review of: ODE-Based Transformer Decoders for Iterative Sign Language Translation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/C7GVKPAW}},
  note         = {Machine review of arXiv:2608.11352}
}
read the original abstract

Sign language translation has achieved strong results with Transformer architectures, yet recent improvements largely rely on scaling model capacity at the cost of increased computation. We propose a parameter-efficient alternative that improves expressiveness without increasing model size. Rather than scaling capacity, we focus on enhancing the update dynamics of iterative refinement decoders, where each refinement step corresponds to one internal decoder iteration that progressively improves the latent representation before translation generation. We reinterpret residual refinement updates from an Ordinary Differential Equation (ODE) perspective and replace them with higher-order numerical integration schemes, namely Runge--Kutta methods (RK-2 and RK-4). These methods perform multiple function evaluations within each refinement step to produce more accurate and stable representation updates without adding decoder parameters. To the best of our knowledge, this is the first application of ODE-inspired update dynamics to sign language translation. RK-2 achieves 22.96 BLEU-4 on the PHOENIX-2014-T test set and 19.34 BLEU-4 on the CSL-Daily test set, outperforming the IPSLT baseline on both benchmarks, with fewer decoder layers and refinement iterations on CSL-Daily. These results suggest that stronger refinement dynamics can improve translation performance under parameter-efficient decoder designs, providing a complementary alternative to conventional model scaling.

Figures

Figures reproduced from arXiv: 2608.11352 by the authors.

Figure 1
Figure 1. Overview of the proposed architecture. A frozen visual encoder extracts fea￾tures, followed by stacked Transformer encoder layers. At each stage, representations are decoded by ODE-based Transformer decoders within an iterative refinement frame￾work. shape, motion, and facial expressions, and capturing long-range dependencies, while the lack of explicit visual–textual alignment and the linguistic differences between… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

59 extracted references · 40 canonical work pages

  1. [1]

    In: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)

    Ahn, N., Kang, B., Sohn, K.A.: Image super-resolution via progressive cascading residual network. In: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). pp. 904–9048 (2018).https://doi.org/10. 1109/CVPRW.2018.00123

    arXiv 2018

  2. [2]

    In: 2021 IEEE/CVF International Conference on Computer Vision (ICCV)

    Alaluf, Y., Patashnik, O., Cohen-Or, D.: Restyle: A residual-based stylegan en- coder via iterative refinement. In: 2021 IEEE/CVF International Conference on Computer Vision (ICCV). pp. 6711–6720 (2021).https://doi.org/10.1109/ ICCV48922.2021.00664 ODE-Based Transformer Decoders for Iterative Sign Language Translation 15

    arXiv 2021

  3. [3]

    Albanie, S., Varol, G., Momeni, L., Bull, H., Afouras, T., Chowdhury, H., Fox, N., Woll, B., Cooper, R., McParland, A., Zisserman, A.: Bbc-oxford british sign language dataset (2021),https://arxiv.org/abs/2111.03635, accessed 8 July 2026

    arXiv 2021

  4. [4]

    In: International Conference on Learning Representations (ICLR) (2025),https://proceedings

    Bae, S., Fisch, A., Harutyunyan, H., Ji, Z., Kim, S., Schuster, T.: Relaxed recursive transformers: Effective parameter sharing with layer-wise lora. In: International Conference on Learning Representations (ICLR) (2025),https://proceedings. iclr.cc/paper_files/paper/2025/hash/54d6a55225cebbdc16fbb0e45c5bdf2b- Abstract-Conference.html

    work page 2025

  5. [5]

    In: Advances in Neural In- formation Processing Systems

    Bae, S., Kim, Y., Bayat, R., Kim, S., Ha, J., Schuster, T., Fisch, A., Harutyunyan, H., Ji, Z., Courville, A.C., Yun, S.Y.: Mixture-of-recursions: Learning dynamic recursive depths for adaptive token-level computation. In: Advances in Neural In- formation Processing Systems. vol. 38 (2025),https://proceedings.neurips. cc / paper _ files / paper / 2025 / h...

    work page 2025

  6. [6]

    In: Advances in Neural Information Processing Systems (NeurIPS) (2019)

    Bai, S., Kolter, J.Z., Koltun, V.: Deep equilibrium models. In: Advances in Neural Information Processing Systems (NeurIPS) (2019)

    work page 2019

  7. [7]

    In: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Camgoz, N.C., Hadfield, S., Koller, O., Ney, H., Bowden, R.: Neural sign language translation. In: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 7784–7793 (2018).https://doi.org/10.1109/CVPR.2018.00812

    arXiv 2018

  8. [8]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Camgoz, N.C., Koller, O., Hadfield, S., Bowden, R.: Sign language transformers: Joint end-to-end sign language recognition and translation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 10023–10033 (2020)

    work page 2020

Show all 59 references
  1. [9]

    In: Proceedings of the 32nd International Conference on Neural Information Processing Systems

    Chen, R.T.Q., Rubanova, Y., Bettencourt, J., Duvenaud, D.: Neural ordinary dif- ferential equations. In: Proceedings of the 32nd International Conference on Neural Information Processing Systems. p. 6572–6583. NIPS’18, Curran Associates Inc., Red Hook, NY, USA (2018)

    2018

  2. [10]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Chen, Y., Wei, F., Sun, X., Wu, Z., Lin, S.: A simple multi-modality transfer learning baseline for sign language translation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 5120–5130 (2022)

    2022

  3. [11]

    In: Proceedings of the 36th Interna- tional Conference on Neural Information Processing Systems

    Chen, Y., Zuo, R., Wei, F., Wu, Y., Liu, S., Mak, B.: Two-stream network for sign language recognition and translation. In: Proceedings of the 36th Interna- tional Conference on Neural Information Processing Systems. NIPS ’22, Curran Associates Inc., Red Hook, NY, USA (2022)

    2022

  4. [12]

    In: Proceedings of the 2024 Joint International Conference on Compu- tational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)

    Chen, Z., Zhou, B., Li, J., Wan, J., Lei, Z., Jiang, N., Lu, Q., Zhao, G.: Fac- torized learning assisted with large language model for gloss-free sign language translation. In: Proceedings of the 2024 Joint International Conference on Compu- tational Linguistics, Language Res...

    2024

  5. [13]

    In: International Conference on Learning Representations (ICLR) (2019), https://openreview.net/forum?id=HyzdRiR9Y7

    Dehghani, M., Gouws, S., Vinyals, O., Uszkoreit, J., Kaiser, L.: Universal trans- formers. In: International Conference on Learning Representations (ICLR) (2019), https://openreview.net/forum?id=HyzdRiR9Y7

    2019

  6. [14]

    In: In- ternational Conference on Learning Representations (ICLR) (2020),https:// openreview.net/forum?id=1KLUy-X5Ayy

    Elbayad, M., Gu, J., Grave, E., Auli, M.: Depth-adaptive transformer. In: In- ternational Conference on Learning Representations (ICLR) (2020),https:// openreview.net/forum?id=1KLUy-X5Ayy

    2020

  7. [15]

    2024 IEEE/CVF Conference on Computer Vision and Pattern Recog- nition (CVPR) pp

    Gong, J., Foo, L.G., He, Y., Rahmani, H., Liu, J.: Llms are good sign language translators. 2024 IEEE/CVF Conference on Computer Vision and Pattern Recog- nition (CVPR) pp. 18362–18372 (2024) 16 T. Kızıltepe and H. Y. Keles

    2024

  8. [16]

    In: Proceedings of the 32nd Inter- national Conference on Machine Learning

    Gregor, K., Danihelka, I., Graves, A., Rezende, D.J., Wierstra, D.: Draw: A re- current neural network for image generation. In: Proceedings of the 32nd Inter- national Conference on Machine Learning. Proceedings of Machine Learning Re- search, vol. 37, pp. 1462–1471 (2015),ht...

    2015

  9. [18]

    In: The Thirty-ninth Annual Conference on Neural Information Processing Systems (2025),https://openreview.net/forum? id=p6Huickfj7, accessed 8 July 2026

    Guo, J., Li, P., Cohn, T.: Bridging sign and spoken languages: Pseudo gloss gen- eration for sign language translation. In: The Thirty-ninth Annual Conference on Neural Information Processing Systems (2025),https://openreview.net/forum? id=p6Huickfj7, accessed 8 July 2026

    2025

  10. [20]

    In: 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Hu, L., Gao, L., Liu, Z., Feng, W.: Continuous sign language recognition with correlation network. In: 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 2529–2539 (2023).https://doi.org/10.1109/ CVPR52729.2023.00249

    2023

  11. [21]

    In: Proceedings of the AAAI Conference on Artificial Intelli- gence

    Hu, L., Gao, L., Liu, Z., Feng, W.: Self-emphasizing network for continuous sign language recognition. In: Proceedings of the AAAI Conference on Artificial Intelli- gence. vol. 37, pp. 854–862 (2023).https://doi.org/10.1609/aaai.v37i1.25164

    2023 doi

  12. [22]

    In: Chiruzzo, L., Ritter, A., Wang, L

    Hwang, E.J., Cho, S., Lee, J., Park, J.C.: An efficient gloss-free sign language translation using spatial configurations and motion dynamics with LLMs. In: Chiruzzo, L., Ritter, A., Wang, L. (eds.) Proceedings of the 2025 Conference of the Nations of the Americas Chapter of t...

    2025 doi

  13. [23]

    ArXiv abs/2512.08040(2025)

    Jang, Y., Momeni, L., Jiang, Z., Chung, J.S., Varol, G., Zisserman, A.: Lost in translation, found in embeddings: Sign language translation and alignment. ArXiv abs/2512.08040(2025)

    2025

  14. [24]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops

    Kızıltepe, T., Taşyürek, S.M., Keles, H.Y.: Iterative latent refinement for robust non-autoregressive sign language production. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops. pp. 4942–4952 (October 2025)

    2025

  15. [25]

    In: Muresan, S., Nakov, P., Villavicencio, A

    Li, B., Du, Q., Zhou, T., Jing, Y., Zhou, S., Zeng, X., Xiao, T., Zhu, J., Liu, X., Zhang, M.: ODE transformer: An ordinary differential equation-inspired model for sequence generation. In: Muresan, S., Nakov, P., Villavicencio, A. (eds.) Proceed- ings of the 60th Annual Meeti...

    2022 doi

  16. [26]

    In: 2025 IEEE/CVF Conference on Computer ODE-Based Transformer Decoders for Iterative Sign Language Translation 17 Vision and Pattern Recognition (CVPR)

    Li, Y., Chen, X., Li, H., Pu, X., Jin, P., Ren, Y.: Vsnet: Focusing on the linguistic characteristics of sign language. In: 2025 IEEE/CVF Conference on Computer ODE-Based Transformer Decoders for Iterative Sign Language Translation 17 Vision and Pattern Recognition (CVPR). pp....

    2025

  17. [27]

    In: International Conference on Learning Representations (ICLR) (2025),https://proceedings.iclr.cc/paper_files/ paper/2025/hash/260a14acce2a89dad36adc8eefe7c59e-Abstract-Conference

    Li, Z., Zhou, W., Zhao, W., Wu, K., Hu, H., Li, H.: Uni-sign: Toward unified sign language understanding at scale. In: International Conference on Learning Representations (ICLR) (2025),https://proceedings.iclr.cc/paper_files/ paper/2025/hash/260a14acce2a89dad36adc8eefe7c59e-A...

    2025

  18. [28]

    In: Text Summarization Branches Out

    Lin, C.Y.: ROUGE: A package for automatic evaluation of summaries. In: Text Summarization Branches Out. pp. 74–81. Association for Computational Linguis- tics, Barcelona, Spain (Jul 2004),https://aclanthology.org/W04- 1013/, ac- cessed 8 July 2026

    2004

  19. [29]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Ling, H., Gao, J., Kar, A., Chen, W., Fidler, S.: Fast interactive object annotation with curve-gcn. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 5257–5266 (2019)

    2019

  20. [30]

    In: 2021 IEEE Winter Conference on Applications of Computer Vision (WACV)

    Liu, Z., Liew, J.H., Chen, X., Feng, J.: Dance : A deep attentive contour model for efficient instance segmentation. In: 2021 IEEE Winter Conference on Applications of Computer Vision (WACV). pp. 345–354 (2021).https://doi.org/10.1109/ WACV48630.2021.00039

    2021

  21. [31]

    In: International Conference on Learning Representations (ICLR) (2019),https://openreview

    Loshchilov, I., Hutter, F.: Decoupled weight decay regularization. In: International Conference on Learning Representations (ICLR) (2019),https://openreview. net/forum?id=Bkg6RiCqY7

    2019

  22. [32]

    In: Thirty-seventh Conference on Neural Information Processing Systems (2023),https://openreview.net/forum?id=S37hOerQLB, accessed 8 July 2026

    Madaan, A., Tandon, N., Gupta, P., Hallinan, S., Gao, L., Wiegreffe, S., Alon, U., Dziri, N., Prabhumoye, S., Yang, Y., Gupta, S., Majumder, B.P., Hermann, K., Welleck, S., Yazdanbakhsh, A., Clark, P.: Self-refine: Iterative refinement with self- feedback. In: Thirty-seventh C...

    2023

  23. [33]

    In: Adjunct Proceedings of the 25th ACM International Conference on Intelligent Virtual Agents

    Mercanoglu Sincan, O., Bowden, R.: Contrastive pretraining with dual visual encoders for gloss-free sign language translation. In: Adjunct Proceedings of the 25th ACM International Conference on Intelligent Virtual Agents. IVA Ad- junct ’25, Association for Computing Machinery...

    2025

  24. [34]

    In: Proceedings of the IEEE/CVF International Con- ference on Computer Vision (ICCV)

    Min, Y., Hao, A., Chai, X., Chen, X.: Visual alignment constraint for continuous sign language recognition. In: Proceedings of the IEEE/CVF International Con- ference on Computer Vision (ICCV). pp. 11542–11551 (October 2021)

    2021

  25. [35]

    In: Isabelle, P., Charniak, E., Lin, D

    Papineni, K., Roukos, S., Ward, T., Zhu, W.J.: Bleu: a method for automatic evaluation of machine translation. In: Isabelle, P., Charniak, E., Lin, D. (eds.) Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics. pp. 311–318. Association for C...

    2002

  26. [36]

    In: Advances in Neural Information Processing Systems

    Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L., Desmaison, A., Köpf, A., Yang, E., DeVito, Z., Raison, M., Tejani, A., Chilamkurthy, S., Steiner, B., Fang, L., Bai, J., Chintala, S.: Pytorch: An imperativ...

    2019

  27. [37]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Peng, S., Jiang, W., Pi, H., Li, X., Bao, H., Zhou, X.: Deep snake for real-time instance segmentation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 8533–8542 (2020) 18 T. Kızıltepe and H. Y. Keles

    2020

  28. [38]

    In: Proceedings of the Third Conference on Machine Translation: Research Papers

    Post, M.: A call for clarity in reporting BLEU scores. In: Proceedings of the Third Conference on Machine Translation: Research Papers. pp. 186–191. Association for Computational Linguistics, Brussels, Belgium (Oct 2018).https://doi.org/10. 18653/v1/W18-6319

    2018

  29. [39]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Ren, D., Zuo, W., Hu, Q., Zhu, P., Meng, D.: Progressive image deraining networks: A better and simpler baseline. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 3937–3946 (2019)

    2019

  30. [40]

    IEEE Transactions on Pattern Analysis and Machine Intelligence45(4), 4713–4726 (2023).https://doi.org/10.1109/TPAMI

    Saharia, C., Ho, J., Chan, W., Salimans, T., Fleet, D.J., Norouzi, M.: Image super- resolution via iterative refinement. IEEE Transactions on Pattern Analysis and Machine Intelligence45(4), 4713–4726 (2023).https://doi.org/10.1109/TPAMI. 2022.3204461

    2023

  31. [41]

    In: Computer Vision – ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23–27, 2022, Proceedings, Part XXIV

    Shen, Z., Liu, Z., Xing, E.: Sliced recursive transformer. In: Computer Vision – ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23–27, 2022, Proceedings, Part XXIV. p. 727–744. Springer-Verlag, Berlin, Heidelberg (2022). https://doi.org/10.1007/978-3-031-20053-3_42

    2022 doi

  32. [42]

    In: Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing

    Shi, B., Brentari, D., Shakhnarovich, G., Livescu, K.: Open-domain sign language translation learned from online video. In: Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing. pp. 6365–6379. Association for Computational Linguistics, Abu Dha...

    2022 doi

  33. [43]

    IEEE Access8, 181340–181355 (2020).https:// doi.org/10.1109/access.2020.3028072

    Sincan, O.M., Keles, H.Y.: Autsl: A large scale multi-modal turkish sign language dataset and baseline methods. IEEE Access8, 181340–181355 (2020).https:// doi.org/10.1109/access.2020.3028072

    2020

  34. [44]

    In: Rambow, O., Wanner, L., Apidi- anaki, M., Al-Khalifa, H., Eugenio, B.D., Schockaert, S

    Tan, S., Miyazaki, T., Khan, N., Nakadai, K.: Improvement in sign language translation using text CTC alignment. In: Rambow, O., Wanner, L., Apidi- anaki, M., Al-Khalifa, H., Eugenio, B.D., Schockaert, S. (eds.) Proceedings of the 31st International Conference on Computational...

    2025

  35. [45]

    Tanzer, G., Zhang, B.: Youtube-sl-25: A large-scale, open-domain multilingual sign language parallel corpus (2024),https://arxiv.org/abs/2407.11144, accessed 8 July 2026

    2024 arXiv

  36. [46]

    In: The Thirteenth International Conference on Learning Representations (2025),https://openreview.net/forum?id=XnDyddPcBT, accessed 8 July 2026

    Tong, A., Nguyen-Tang, T., Lee, D., Nguyen, D., Tran, T., Hall, D.L.W., Kang, C., Choi,J.:NeuralODEtransformers:Analyzinginternaldynamicsandadaptivefine- tuning. In: The Thirteenth International Conference on Learning Representations (2025),https://openreview.net/forum?id=XnDy...

    2025

  37. [47]

    In: Advances in Neural Information Pro- cessing Systems

    Uthus, D., Tanzer, G., Georg, M.: Youtube-asl: A large-scale, open-domain ameri- can sign language-english parallel corpus. In: Advances in Neural Information Pro- cessing Systems. vol. 36, pp. 29029–29047 (2023),https://proceedings.neurips. cc / paper _ files / paper / 2023 /...

    2023

  38. [48]

    In: Proceedings of the IEEE/CVF International Conference on Com- puter Vision (ICCV)

    Wei, F., Chen, Y.: Improving continuous sign language recognition with cross- lingual signs. In: Proceedings of the IEEE/CVF International Conference on Com- puter Vision (ICCV). pp. 23612–23621 (2023)

    2023

  39. [49]

    In: The Twelfth International Confer- ence on Learning Representations (2024),https://openreview.net/forum?id= LqaEEs3UxU, accessed 8 July 2026

    Wong, R., Camgoz, N.C., Bowden, R.: Sign2GPT: Leveraging large language mod- els for gloss-free sign language translation. In: The Twelfth International Confer- ence on Learning Representations (2024),https://openreview.net/forum?id= LqaEEs3UxU, accessed 8 July 2026

    2024

  40. [50]

    In: 2026 IEEE ODE-Based Transformer Decoders for Iterative Sign Language Translation 19 International Conference on Acoustics, Speech and Signal Processing (ICASSP)

    Wu, W., Yuan, T., Li, Y., Wang, D., Fu, X.: Mouthing-enhanced multimodal hier- archical contrastive learning for gloss-free sign language translation. In: 2026 IEEE ODE-Based Transformer Decoders for Iterative Sign Language Translation 19 International Conference on Acoustics,...

    2026

  41. [51]

    Xu, R., Gao, Y., Wang, L., Li, J., Chen, W., Guo, Q., Yang, M., Zhang, S.: Loop- ing back to move forward: Recursive transformers for efficient and flexible large multimodal models (2026),https://arxiv.org/abs/2602.09080, accessed 8 July 2026

    2026

  42. [52]

    In: Proceedings of the IEEE/CVF International Confer- ence on Computer Vision (ICCV)

    Yao, H., Zhou, W., Feng, H., Hu, H., Zhou, H., Li, H.: Sign language translation with iterative prototype. In: Proceedings of the IEEE/CVF International Confer- ence on Computer Vision (ICCV). pp. 15592–15601 (2023)

    2023

  43. [53]

    In: Bouamor, H., Pino, J., Bali, K

    Ye, J., Jiao, W., Wang, X., Tu, Z., Xiong, H.: Cross-modality data augmenta- tion for end-to-end sign language translation. In: Bouamor, H., Pino, J., Bali, K. (eds.) Findings of the Association for Computational Linguistics: EMNLP 2023. pp.13558–13571.AssociationforComputatio...

    2023 doi

  44. [54]

    In: Advances in Neural Information Processing Systems

    Ye, J., Wang, X., Jiao, W., Liang, J., Xiong, H.: Improving gloss-free sign lan- guage translation by reducing representation density. In: Advances in Neural Information Processing Systems. vol. 37 (2024).https://doi.org/10.52202/ 079017- 3411,https://proceedings.neurips.cc/pa...

    2024

  45. [55]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Yin, A., Zhong, T., Tang, L., Jin, W., Jin, T., Zhao, Z.: Gloss attention for gloss- free sign language translation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 2551–2562 (2023)

    2023

  46. [56]

    In: The Eleventh International Conference on Learning Rep- resentations (2023),https://openreview.net/forum?id=EBS4C77p_5S, accessed 8 July 2026

    Zhang, B., Müller, M., Sennrich, R.: SLTUNET: A simple unified model for sign language translation. In: The Eleventh International Conference on Learning Rep- resentations (2023),https://openreview.net/forum?id=EBS4C77p_5S, accessed 8 July 2026

    2023

  47. [57]

    In: Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition (CVPR)

    Zhang, T., Wei, S., Ji, S.: E2ec: An end-to-end contour-based method for high- quality high-speed instance segmentation. In: Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition (CVPR). pp. 4443–4452 (2022)

    2022

  48. [58]

    In: The Symbiosis of Deep Learning and Differential Equations II (2022),https://openreview.net/forum?id=nA9hvYMQCy, accessed 8 July 2026

    Zhong, Y.D., Zhang, T., Chakraborty, A., Dey, B.: A neural ODE interpretation of transformer layers. In: The Symbiosis of Deep Learning and Differential Equations II (2022),https://openreview.net/forum?id=nA9hvYMQCy, accessed 8 July 2026

    2022

  49. [59]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)

    Zhou, B., Chen, Z., Clapés, A., Wan, J., Liang, Y., Escalera, S., Lei, Z., Zhang, D.: Gloss-free sign language translation: Improving from visual-language pretraining. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). pp. 20871–20881 (October 2023)

    2023

  50. [60]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Zhou, H., Zhou, W., Qi, W., Pu, J., Li, H.: Improving sign language translation with monolingual data by sign back-translation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 1316–1325 (2021)

    2021

  51. [61]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Zuo, R., Wei, F., Mak, B.: Natural language-assisted sign language recognition. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 14890–14900 (2023)

    2023

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.