Pith. sign in

REVIEW 2 major objections 5 minor 46 references

Q-BridgeNet: A Quantization Network for Cross-Lingual Sign Language Translation

T0 review · 2 major / 5 minor · reviewed 2026-07-14 · grok-4.5

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

Pith's one-line read A single model can translate American, Chinese and German sign languages into English, Chinese and German by turning continuous signing into shared discrete Q-units that a multilingual language model reads.

desk verdict Solid native SOTA via adaptive shared–private residual Q-units; the many-to-many story is real but partly GPT-caption-dependent. read the letter →

arxiv 2607.11215 v1 pith:OTKEHL35 submitted 2026-07-13 cs.CL cs.MM

Liqian Feng , Lintao Wang , Xiaochen Liu , Anusha Withana , Ken-Tye Yong , Dehui Kong , Zhiyong Wang , Kun Hu This is my paper · ORCID
classification cs.CLcs.MM
keywords signlanguagetranslationmultilingualmodelingvectorquantizationresidualcodebooksadaptivesegmentationlargemodelscross-lingualtransfer
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

Most sign-language translation systems only handle one sign language paired with its local spoken language. This paper argues that many-to-many translation becomes practical once continuous pose streams are turned into discrete Q-units: variable-length segments that are quantized with a shared base codebook for language-agnostic meaning and language-specific residual codebooks for local signing style. A multilingual language model is then fine-tuned to map those Q-units into any of the target spoken languages. On the three standard datasets How2Sign, CSL-Daily and PHOENIX14T the method reports state-of-the-art native-pair scores and strong transfer to non-native pairs. If the claim holds, people using different sign languages could communicate through one system instead of a patchwork of pair-specific models.

What carries the argument

Discrete Q-units: adaptive variable-length segmentation of pose sequences, jointly refined with multi-codebook residual vector quantization (shared base codebook plus language-specific residual codebooks), after which a multilingual LLM is fine-tuned to operate directly in the resulting Q-unit token space.

What would settle it

Retrain the identical pipeline using only human-authored multilingual captions (or evaluate non-native outputs against a held-out set of human translations) and check whether native and non-native BLEU/ROUGE still exceed the strongest prior baselines by the same margins reported in Table 1.

Watch

Extended reading notes

Core claim

Residual vector quantization of adaptively segmented pose streams—using one shared base codebook for cross-lingual semantic primitives and private residual codebooks for language-specific detail—produces discrete Q-units that reduce cross-lingual interference while remaining expressive enough for a fine-tuned multilingual LLM to reach state-of-the-art native sign-to-text scores and strong non-native generalization on PHOENIX14T, How2Sign and CSL-Daily.

Load-bearing premise

The method treats machine translations of the original captions as both training targets and evaluation references for non-native pairs; if those translations systematically simplify or style-match the decoder, measured cross-lingual gains partly reflect caption artifacts rather than true sign understanding.

Editorial extensions

If this is right

  • A single many-to-many SLT model can replace separate monolingual systems for ASL–English, CSL–Chinese and DGS–German.
  • Non-native pairs (for example ASL to Chinese) become usable without re-learning sign units for each spoken language.
  • The same discrete Q-unit inventory can later serve as the bridge for the reverse task of sign language production.
  • Shared–private residual quantization can limit negative transfer when additional sign languages are added to the model.

Reading between the lines

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

  • The alternating segmentation–quantization loop may transfer to other continuous multimodal streams whose natural temporal boundaries are ambiguous.
  • If Q-units truly encode reusable semantic primitives, retrieval-augmented or few-shot cross-lingual SLT could become cheaper to scale.
  • Stronger human-translated multilingual corpora would be a direct next test of whether the reported non-native gains are limited by caption quality.
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

2 major / 5 minor

Summary. Q-BridgeNet proposes a unified many-to-many sign language translation framework that represents continuous pose streams as discrete Q-units obtained by alternating adaptive variable-length segmentation with residual vector quantization. A shared base codebook is intended to capture language-agnostic signing primitives, while language-specific residual codebooks refine ASL/CSL/DGS variation; a multilingual LLM (Qwen3-1.7B with LoRA) is then fine-tuned to map Q-unit sequences to spoken text in German, English, and Chinese. Multilingual supervision is created by translating native captions with GPT-5. On PHOENIX14T, How2Sign, and CSL-Daily the method reports state-of-the-art native BLEU/ROUGE and competitive non-native scores (Table 1), with ablations of codebook sharing, residual stacking, iterative segmentation, and LLM scale (Table 2) and supporting t-SNE/gloss-frequency analyses of the discrete space.

Significance. If the claims hold, the work would be a useful step toward practical multilingual SLT: a single model that both improves native pairs and transfers across sign–spoken combinations without per-pair re-tokenization. The shared–private residual design and the joint segmentation–quantization procedure are concrete technical contributions relative to fixed-length single-language VQ and continuous multilingual baselines such as Uni-Sign. Public code, systematic ablations (Table 2), and distributional analyses of Q-units (Figs. 4–5) strengthen reproducibility and interpretability. The main scientific stake is whether the non-native gains reflect transferable sign semantics rather than properties of machine-translated captions used for both training and evaluation.

major comments (2)
  1. §3.4 and §4.1 Evaluation Metrics: non-native training targets and non-native evaluation references are both produced by GPT-5 (or GPT-4.1) from the same native captions. The central claim of “strong generalization to non-native pairs” (abstract, §4.3, Table 1 rows ASL-ZH/DE, CSL-EN/DE, DGS-EN/ZH) therefore rests on machine-translated references. The Table 2 “GPT-5 Translation” control (monolingual SLT then post-hoc translate) is lower and helpful, but does not retrain under independent non-GPT multilingual supervision nor evaluate against human non-native references. Without at least a human-validated subset or an independent MT system for references, measured non-native BLEU/ROUGE may partly reflect caption style/simplification rather than cross-lingual sign modeling. This is load-bearing for the many-to-many contribution even though native SOTA (original GT) is less affected.
  2. §3.3, Eqs. (7)–(13): the joint objective J(S,V) and the block-coordinate alternation are presented as yielding “semantically coherent” variable-length units, yet the only quantitative evidence is reconstruction-driven segmentation plus downstream BLEU. There is no direct boundary-quality metric (e.g., agreement with gloss boundaries where available, or held-out reconstruction under fixed vs. adaptive segments). Given that adaptive segmentation is a core claimed advantage over fixed-length VQ, a load-bearing validation of segment quality—or a clearer statement that semantic coherence is only inferred from SLT gains—would strengthen the central representation claim.
minor comments (5)
  1. Table 1: “Avg.” rows for the multilingual model are not defined (mean over three targets? weighted?). Clarify how averages are computed and whether they are comparable across datasets of different sizes.
  2. §3.2: residual depth R, exact codebook sizes (1024 shared / 512 private), and the factor-binning of residual codebooks are stated in Implementation Details but not fully formalized in the equations; a short explicit definition of language-conditioned Q^r_i would help.
  3. Fig. 3 qualitative examples: bold/strike markup is useful, but adding the corresponding Q-unit sequence length or segment count would make the link to adaptive segmentation more concrete.
  4. Related Work: Uni-Sign and MGSLT are discussed; a short explicit contrast on discrete vs. continuous interfaces and on whether those works also use MT-augmented targets would sharpen novelty.
  5. Typos/spacing: “SL T” in Fig. 1 caption; occasional missing spaces after periods in the abstract and §1; “factor-binned” could be defined once for non-VQ readers.

Circularity Check

0 steps flagged · score 2.0 of 10

No derivation circularity; native SOTA uses original GT while non-native GPT loop is a disclosed evaluation choice with control, not a by-construction reduction of the method.

full rationale

Q-BridgeNet is an empirical systems paper whose core claims rest on adaptive variable-length segmentation + shared-base/language-specific residual VQ (Eqs. 1-13, joint objective J) followed by LLM fine-tuning on the resulting discrete tokens (Eq. 15). These steps are self-contained algorithmic constructions optimized against reconstruction and next-token losses; they do not reduce any reported quantity to its own inputs by definition, nor invoke self-citation uniqueness theorems or smuggled ansatze. Native-pair SOTA (Table 1 ♣ rows) is measured against the original human ground-truth of each benchmark and is therefore independent of GPT. The many-to-many non-native numbers use GPT-5 (and GPT-4.1) translations both as training targets (§3.4) and as references (§4.1), which creates a shared synthetic distribution; however this is a methodological evaluation choice, not a fitted parameter renamed as prediction, and the paper supplies an explicit lower-scoring post-hoc GPT-5 Translation control (Table 2). No equation or uniqueness claim collapses by construction. Score remains low (honest non-finding for derivation circularity).

Assumptions & free parameters 8 free parameters · 5 assumptions · 2 invented entities

The central performance claims rest on standard VQ-VAE/LLM machinery plus several paper-specific modeling choices and free hyperparameters (codebook sizes, residual structure, segmentation regularizer, GPT-augmented labels). Q-units are the main invented interface entity; their semantic status is supported only by reconstruction, ablations, t-SNE, and gloss-frequency resemblance, not by independent linguistic validation.

free parameters (8)
  • shared base codebook size = 1024
    Set to 1,024 embeddings; capacity of language-agnostic primitives is a design choice that directly affects reported transfer.
  • language-specific residual codebook size = 512
    Private codebooks of 512 embeddings per language bin; controls how much variation is treated as language-specific.
  • VQ commitment weights β, γ = β=1, γ=0.25
    β=1, γ=0.25 chosen empirically in the RQ-VAE loss; alter codebook usage and reconstruction trade-off.
  • max segment length / max sequence length = 32 / 256
    Preliminary max segment 32 frames and T=256 cap the temporal units available to quantization and the LLM.
  • latent dimension and residual depth R = 96-d latent; R residual layers
    96-d latents and R residual layers (exact R not fully stressed in main text) define the discrete token geometry.
  • segmentation count regularizer λ and alternation rounds = 5 alternation rounds; λ unspecified numeric
    λN and 5 block-coordinate iterations (plus ε stopping) shape how many Q-units are produced; not derived from theory.
  • keypoint noise std and pose topology = σ=0.002; 79 keypoints
    79 keypoints and i.i.d. Gaussian noise σ=0.002 during training are hand-set robustness knobs.
  • LLM backbone and LoRA fine-tuning schedule = Qwen3-1.7B, 100 epochs
    Qwen3-1.7B, LoRA, 100 epochs, prompt templates; capacity and optimization choices affect absolute BLEU.
assumptions (5)
  • domain assumption Different sign languages share reusable language-agnostic semantic primitives that a single base codebook can capture while residuals absorb language-specific realization.
    Stated as the key design idea in the introduction and §3.2; not independently proven, only motivated and probed via ablations/t-SNE.
  • ad hoc to paper Minimizing cumulative RQ reconstruction error with alternating segmentation yields temporally and semantically coherent signing units suitable as translation tokens.
    Joint objective J(S,V) and block-coordinate descent in §3.3 treat reconstruction coherence as a proxy for semantic units.
  • ad hoc to paper Machine-translated spoken captions (GPT-5/4.1) are adequate many-to-many supervision and, for non-native pairs, adequate evaluation references.
    §3.4 constructs trilingual X via GPT-5; §4.1 evaluates non-native pairs against those translations.
  • domain assumption Skeleton/pose sequences (hands–body–face landmarks) preserve enough articulatory information for competitive SLT without appearance features.
    Implementation Details adopt 79 keypoints following common SLT practice; limits what the model can see.
  • standard math Standard residual VQ-VAE training (reconstruction + commitment losses) and next-token LLM fine-tuning are valid optimization targets for discrete sign–text alignment.
    Equations (5) and (15) import standard VQ-VAE and autoregressive LM losses.
invented entities (2)
  • Q-units (variable-length residual quantized signing tokens)
    purpose: Serve as the shared discrete interface between multilingual signing streams and a single LLM for many-to-many SLT.
    Defined via adaptive segmentation + shared-base residual VQ; independent linguistic status is only weakly evidenced (t-SNE clusters, gloss frequency resemblance), not by external annotation agreement or psycholinguistic tests.
  • Shared–private residual codebook factorization for sign languages (shared Z0 + language-binned residual Zr)
    purpose: Separate cross-lingual primitives from language-specific refinements to reduce cross-lingual interference.
    Architectural postulate of §3.2; support is ablation performance, not an external measurement of interference reduction.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Q-BridgeNet: A Quantization Network for Cross-Lingual Sign Language Translation." pith.science (2026). https://pith.science/paper/OTKEHL35

@misc{pith2026260711215,
  author       = {Pith},
  title        = {Pith review of: Q-BridgeNet: A Quantization Network for Cross-Lingual Sign Language Translation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OTKEHL35}},
  note         = {Machine review of arXiv:2607.11215}
}
read the original abstract

Most sign language translation (SLT) methods focus on isolated native sign-spoken pairs (e.g., American Sign Language - English). Extending language-specific SLT models to multilingual translation would improve accessibility by enabling communication across diverse sign and spoken language communities. However, existing multilingual SLT approaches still struggle to learn a unified model that minimizes cross-lingual conflicts while capturing shared cross-lingual semantics and preserving language-specific variations across different sign languages. Therefore, we propose Q-BridgeNet, a unified framework for multilingual SLT that jointly mitigates cross-lingual conflicts across both the sign language and spoken language sides. On the sign language side, Q-BridgeNet learns discrete Q-units via adaptive segmentation and residual vector quantization: a shared base codebook provides language-agnostic semantic primitives, while language-specific residual codebooks refine heterogeneous signing semantics. On the spoken language side, a multilingual LLM is fine-tuned to operate in the Q-unit space, leveraging cross-lingual priors to enable a unified SLT model. Experiments on PHOENIX14T, How2Sign, and CSL-Daily show that Q-BridgeNet effectively mitigates cross-lingual conflicts, achieving state-of-the-art performance on native sign-spoken pairs while also demonstrating strong generalization to non-native pairs. Our source code is publicly available at: https://github.com/FengLiQ/Q-BridgeNet

Figures

Figures reproduced from arXiv: 2607.11215 by the authors.

Figure 1
Figure 1. Multilingual SLT and signing units. (a) Monolingual SLT translates a sin￾gle sign–spoken language pair. (b) Multilingual SLT extends to many-to-many transla￾tion across multiple sign and spoken languages. (c) Fixed-length segmentation may cut across semantic/lexical boundaries, mixing heterogeneous signing patterns and yield￾ing poor alignment to text. (d) Variable-length semantic Q-Unit forms coherent signing units… view at source ↗
Figure 2
Figure 2. Q-BridgeNet Overview. (a) Iterative Q-Unit Discovery: Segmentation alter￾nates with residual VQ to produce variable-length discrete sign tokens from pose se￾quences. A shared base codebook captures language-agnostic primitives, while language￾specific residual codebooks (ASL/CSL/DGS) add semantic detail. (b) Multilingual LLM Re-Alignment: A pretrained LLM is adapted to operate in the Q-unit space, en￾abling a unifie… view at source ↗
Figure 3
Figure 3. Multilingual qualitative examples for SLT. Bold: semantically aligned with GT; and Strike: missing semantics. In native settings, the model produces fluent and accurate translations that align closely with the ground-truth text, effectively leveraging both the shared base codebook and language-specific residual codebooks to handle linguistic vari￾ations. In non-native scenarios—where sign languages are paired with a… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: t-SNE visualization of discrete sign tokens. (a) Fixed-length segmentation with a single VQ-VAE results in scattered and semantically entangled clusters. (b) Our variable-length segmentation with three language-specific codebooks (VASL, VCSL, and VDGS) yields well-sepa…
Figure 5
Figure 5. Figure 5: Comparison between gloss frequency distributions (blue) and private Q-unit token distributions (red) on CSL and DGS. The similar long-tailed trends indicate structural resemblance between learned Q-units and linguistic-level annotation units. 4.6 Analysis of Q-unit Spa…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references

  1. [1]

    In: IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (2018)

    Camgoz, N.C., Hadfield, S., Koller, O., Ney, H., Bowden, R.: Neural sign language translation. In: IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (2018)

    2018

  2. [2]

    In: IEEE/CVF Con- ference on Computer Vision and Pattern Recognition (CVPR) (2020)

    Camgoz, N.C., Koller, O., Hadfield, S., Bowden, R.: Sign language transformers: Joint end-to-end sign language recognition and translation. In: IEEE/CVF Con- ference on Computer Vision and Pattern Recognition (CVPR) (2020)

    2020

  3. [3]

    IEEE Transactions on Pattern Analysis and Machine Intelligence (2021)

    Cao, Z., Hidalgo, G., Simon, T., Wei, S.E., Sheikh, Y.: Openpose: Realtime multi- person 2d pose estimation using part affinity fields. IEEE Transactions on Pattern Analysis and Machine Intelligence (2021)

    2021

  4. [4]

    In: International Conference on Neural Information Processing Systems (NeurIPS) (2022)

    Chen, Y., Zuo, R., Wei, F., Wu, Y., Liu, S., Mak, B.: Two-stream network for sign language recognition and translation. In: International Conference on Neural Information Processing Systems (NeurIPS) (2022)

    2022

  5. [5]

    IEEE Transactions on Circuits and Systems for Video Technology (2025)

    Chen, Z., Zhou, B., Huang, Y., Wan, J., Hu, Y., Shi, H., Liang, Y., Lei, Z., Zhang, D.: C2rl: Content and context representation learning for gloss-free sign language translation and retrieval. IEEE Transactions on Circuits and Systems for Video Technology (2025)

    2025

  6. [6]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2021)

    Duarte, A., Palaskar, S., Ventura, L., Ghadiyaram, D., DeHaan, K., Metze, F., Torres, J., Giro-i Nieto, X.: How2sign: a large-scale multimodal dataset for contin- uous american sign language. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2021)

    2021

  7. [7]

    In: Interna- tional Conference on Language Resources and Evaluation (LREC) (2014)

    Forster, J., Schmidt, C., Koller, O., Bellgardt, M., Ney, H.: Extensions of the sign language recognition and translation corpus rwth-phoenix-weather. In: Interna- tional Conference on Language Resources and Evaluation (LREC) (2014)

    2014

  8. [8]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion (CVPR) (2024)

    Gong, J., Foo, L.G., He, Y., Rahmani, H., Liu, J.: Llms are good sign language translators. In: IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion (CVPR) (2024)

    2024

Show all 46 references
  1. [9]

    Pattern Recognition (2025)

    Guan, M., Wang, Y., Ma, G., Liu, J., Sun, M.: Mska: Multi-stream keypoint at- tention network for sign language recognition and translation. Pattern Recognition (2025)

    2025

  2. [10]

    In: Annual Meeting of the Association for Computational Linguistics (ACL) (2025)

    Gueuwou, S., Du, X., Shakhnarovich, G., Livescu, K., Liu, A.H.: Shubert: Self- supervised sign language representation learning via multi-stream cluster predic- tion. In: Annual Meeting of the Association for Computational Linguistics (ACL) (2025)

    2025

  3. [11]

    In: International Conference on Learning Representations (ICLR) (2022)

    Hu, E.J., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., Wang, L., Chen, W.: Lora: Low-rank adaptation of large language models. In: International Conference on Learning Representations (ICLR) (2022)

    2022

  4. [12]

    In: AAAI Conference on Artificial Intelligence (AAAI) (2018)

    Huang, J., Zhou, W., Zhang, Q., Li, H., Li, W.: Video-based sign language recogni- tion without temporal segmentation. In: AAAI Conference on Artificial Intelligence (AAAI) (2018)

    2018

  5. [13]

    In: Annual Meeting of the Association for Computational Linguistics (ACL) (2025)

    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: Annual Meeting of the Association for Computational Linguistics (ACL) (2025)

    2025

  6. [14]

    In: Advances in Neural Information Processing Systems (NeurIPS) (2023)

    Jiang, B., Chen, X., Liu, W., Yu, J., Yu, G., Chen, T.: Motiongpt: Human motion as a foreign language. In: Advances in Neural Information Processing Systems (NeurIPS) (2023)

    2023

  7. [15]

    In: European Conference on Computer Vision (ECCV) (2024) Q-BridgeNet: A Quantization Network for Cross-Lingual SLT 17

    Jiao, P., Min, Y., Chen, X.: Visual alignment pre-training for sign language trans- lation. In: European Conference on Computer Vision (ECCV) (2024) Q-BridgeNet: A Quantization Network for Cross-Lingual SLT 17

    2024

  8. [16]

    In: IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) (2022)

    Kan, J., Hu, K., Hagenbuchner, M., Tsoi, A.C., Bennamoun, M., Wang, Z.: Sign language translation with hierarchical spatio-temporal graph neural network. In: IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) (2022)

    2022

  9. [17]

    In: IEEE/CVF International Conference on Computer Vision (ICCV) (2025)

    Kim, J., Jeon, H., Bae, J., Kim, H.Y.: Leveraging the power of mllms for gloss-free sign language translation. In: IEEE/CVF International Conference on Computer Vision (ICCV) (2025)

    2025

  10. [18]

    In: International Conference on Learning Representations (ICLR) (2025)

    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)

    2025

  11. [19]

    In: Text Sum- marization Branches Out (2004)

    Lin, C.Y.: Rouge: A package for automatic evaluation of summaries. In: Text Sum- marization Branches Out (2004)

    2004

  12. [20]

    Lin, K., Wang, X., Zhu, L., Sun, K., Zhang, B., Yang, Y.: Gloss-free end-to-end signlanguagetranslation.In:AnnualMeetingoftheAssociationforComputational Linguistics (ACL) (2023)

    2023

  13. [21]

    In: AAAI Conference on Artificial Intelligence (AAAI) (2025)

    Liu, Y., Zhang, W., Ren, S., Huang, C., Yu, J., Xu, L.: Scope: Sign language contextual processing with embedding from llms. In: AAAI Conference on Artificial Intelligence (AAAI) (2025)

    2025

  14. [22]

    Journal of Machine Learn- ing Research (2008)

    Maaten, L.v.d., Hinton, G.: Visualizing data using t-sne. Journal of Machine Learn- ing Research (2008)

    2008

  15. [23]

    Language Learning and Development (2010)

    Mann, W., Marshall, C.R., Mason, K., Morgan, G.: The acquisition of sign lan- guage: The impact of phonetic complexity on phonology. Language Learning and Development (2010)

    2010

  16. [24]

    Expert Systems with Applications (2023)

    Núñez-Marcos, A., de Viñaspre, O.P., Labaka, G.: A survey on sign language ma- chine translation. Expert Systems with Applications (2023)

    2023

  17. [25]

    OpenAI: Introducing GPT-4.1.https://openai.com/index/gpt- 4- 1/(April 2025), accessed: 2025-4-16

    2025

  18. [26]

    OpenAI: Introducing GPT-5.https://openai.com/index/introducing-gpt-5/ (August 2025), accessed: 2025-8-10

    2025

  19. [27]

    In: Annual Meeting of the Association for Com- putational Linguistics (ACL) (2002)

    Papineni, K., Roukos, S., Ward, T., Zhu, W.J.: Bleu: a method for automatic evaluation of machine translation. In: Annual Meeting of the Association for Com- putational Linguistics (ACL) (2002)

    2002

  20. [28]

    In: Conference on Machine Translation: Research Papers (2018)

    Post, M.: A call for clarity in reporting BLEU scores. In: Conference on Machine Translation: Research Papers (2018)

    2018

  21. [29]

    In: AAAI Conference on Artificial Intelligence (AAAI) (2025)

    Shen, F., Jiang, X., He, X., Ye, H., Wang, C., Du, X., Li, Z., Tang, J.: Imagdressing- v1: Customizable virtual dressing. In: AAAI Conference on Artificial Intelligence (AAAI) (2025)

    2025

  22. [30]

    Advances in neural information processing systems (2024)

    Shen, F., Tang, J.: Imagpose: A unified conditional framework for pose-guided person generation. Advances in neural information processing systems (2024)

    2024

  23. [31]

    In: IEEE/CVF International Conference on Computer Vision (CVPR) (2026)

    Shi, C., Li, S., Guo, S., Xie, S., Wu, W., Dou, J., Wu, C., Xiao, C., Wang, C., Cheng, Z., Shen, F., Chua, T.S.: Where culture fades: Revealing the cultural gap in text-to-image generation. In: IEEE/CVF International Conference on Computer Vision (CVPR) (2026)

    2026

  24. [32]

    In: International Conference on Computational Linguistics (COLING) (2025)

    Tan, S., Miyazaki, T., Khan, N., Nakadai, K.: Improvement in sign language trans- lation using text ctc alignment. In: International Conference on Computational Linguistics (COLING) (2025)

    2025

  25. [33]

    In: Annual Meeting of the Association for Computational Linguistics (ACL) (2025) 18 L

    Tan, S., Miyazaki, T., Nakadai, K.: Multilingual gloss-free sign language transla- tion: Towards building a sign language foundation model. In: Annual Meeting of the Association for Computational Linguistics (ACL) (2025) 18 L. Feng et al

    2025

  26. [34]

    In: Advances in Neural Information Processing Systems (NeurIPS) (2023)

    Uthus, D., Tanzer, G., Georg, M.: Youtube-asl: A large-scale, open-domain amer- ican sign language-english parallel corpus. In: Advances in Neural Information Processing Systems (NeurIPS) (2023)

    2023

  27. [35]

    In: IEEE International Conference on Automatic Face and Gesture Recognition (2024)

    Walsh, H., Ravanshad, A., Rahmani, M., Bowden, R.: A data-driven representation for sign language production. In: IEEE International Conference on Automatic Face and Gesture Recognition (2024)

    2024

  28. [36]

    In: AAAI Conference on Artificial Intelligence (AAAI) (2025)

    Wu, Q., Huang, H., Su, K., Wang, Z., Hu, K.: Dc-pcn: Point cloud completion net- work with dual-codebook guided quantization. In: AAAI Conference on Artificial Intelligence (AAAI) (2025)

    2025

  29. [37]

    In: IEEE/CVF conference on computer vision and pattern recognition (CVPR) (2022)

    Yin, A., Zhao, Z., Jin, W., Zhang, M., Zeng, X., He, X.: Mlslt: Towards multilin- gual sign language translation. In: IEEE/CVF conference on computer vision and pattern recognition (CVPR) (2022)

    2022

  30. [38]

    In: European Conference on Computer Vision (ECCV) (2024)

    Yu,Z.,Huang,S.,Cheng,Y.,Birdal,T.:Signavatars:Alarge-scale3dsignlanguage holistic motion dataset and benchmark. In: European Conference on Computer Vision (ECCV) (2024)

    2024

  31. [39]

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

    Zhang, J., Zhang, Y., Cun, X., Zhang, Y., Zhao, H., Lu, H., Shen, X., Shan, Y.: Generating human motion from textual descriptions with discrete representations. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2023)

    2023

  32. [40]

    In: International Conference on Computing and Pattern Recognition (ICCPR) (2025)

    Zhang, L., Liu, Y., Luo, Y., Gao, F., Gu, J.: Qwen-ig: A qwen-based instruction generation model for llm fine-tuning. In: International Conference on Computing and Pattern Recognition (ICCPR) (2025)

    2025

  33. [41]

    In: International Conference on Computational Linguistics (COLING) (2025)

    Zhang, R., Hu, C., Yu, P., Chen, Y.: Improving multilingual sign language trans- lation with automatically clustered language family information. In: International Conference on Computational Linguistics (COLING) (2025)

    2025

  34. [42]

    In: AAAI Conference on Artificial Intelligence (AAAI) (2024)

    Zhao, R., Zhang, L., Fu, B., Hu, C., Su, J., Chen, Y.: Conditional variational autoencoder for sign language translation with cross-modal alignment. In: AAAI Conference on Artificial Intelligence (AAAI) (2024)

    2024

  35. [43]

    In: IEEE/CVF International Conference on Computer Vision (ICCV) (2023)

    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: IEEE/CVF International Conference on Computer Vision (ICCV) (2023)

    2023

  36. [44]

    IEEE Transactions on Pattern Analysis and Machine Intelligence (2025)

    Zhou, W., Zhao, W., Hu, H., Li, Z., Li, H.: Scaling up multimodal pre-training for sign language understanding. IEEE Transactions on Pattern Analysis and Machine Intelligence (2025)

    2025

  37. [45]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    Zhou, Z., Wan, Y., Wang, B.: Avatargpt: All-in-one framework for motion under- standing planning generation and beyond. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    2024

  38. [46]

    In: IEEE/CVF Interna- tional Conference on Computer Vision (ICCV) (2025)

    Zuo, R., Potamias, R.A., Ververas, E., Deng, J., Zafeiriou, S.: Signs as tokens: A retrieval-enhanced multilingual sign language generator. In: IEEE/CVF Interna- tional Conference on Computer Vision (ICCV) (2025)

    2025

Pith tools

Reviewed July 14, 2026 · model on record in the stance chip above.