Pith. sign in

REVIEW 3 major objections 2 minor 35 references

NAVER LABS Europe Submission to the Instruction-following Track

T0 review · 3 major / 2 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A speech projector and text LoRA adapters, trained apart, fuse in 1,000 joint steps into a model that beats speech-only systems on ASR, ST, and spoken QA.

desk verdict The parallel-projector-plus-LoRA recipe is genuinely useful, but the SQA numbers and the Seamless comparison need to be fixed before I trust the headline claims. read the letter →

arxiv 2506.01808 v1 pith:MCTK5QOX submitted 2025-06-02 cs.CL

Beomseok Lee , Marcely Zanon Boito , Laurent Besacier , Ioan Calapodescu This is my paper
classification cs.CL
keywords speechinstructionfollowingspokenquestionansweringtranslationprojectorLoRAadaptersjointtuningmultimodallargelanguagemodelsASR
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

This paper is a system description for a constrained multilingual speech instruction-following track: one model must recognize English speech, translate it into German, Italian, or Chinese, and answer questions about the spoken content in those languages. The claim it argues for is that the speech and text sides of such a model do not need to be trained jointly from the start. A speech-to-LLM embedding projector (A) is trained on speech-to-text data while text-only LoRA adapters (B) are trained on text-to-text data in parallel. A short 1,000-step joint instruction-tuning stage (C) then aligns the two, and the fused model improves ASR and translation over the speech-projector-only system and, on spoken QA, approaches the text-only baseline. The practical interest is that the expensive per-modality training is decoupled and the merge is cheap, so hyperparameter effort can concentrate on the final alignment stage.

What carries the argument

The load-bearing pieces are three. (1) A speech projector: a four-layer transformer encoder that maps averaged speech representations from a large pretrained multilingual speech encoder into the embedding space of a frozen instruction-tuned 8-billion-parameter LLM, letting the LLM consume audio as if it were text. (2) Text LoRA adapters: low-rank updates to the attention and feed-forward weights of the same frozen LLM, trained on text-only translation and QA data. (3) The C-stage alignment: loading both modules and running 1K instruction-tuning steps over interleaved speech and text batches, with the textual batches acting as a stabilizing signal; the authors ablate that a randomly initialized LoRA gives less gain than the pretrained text LoRA. Averaging every third frame of the speech encoder output also shortens sequences so larger batches fit in memory.

What would settle it

Score the submitted A.1+B fluent model's outputs on the multilingual SpokenSQuAD test set with the organizers' BERT-score settings; if the en-de, en-it, and en-zh BERT scores remain near the reported 0.532-0.666 range while the LLM judge reports 76-85% accuracy, the SQA claim fails under the official metric.

Watch

Extended reading notes

Core claim

The paper's central discovery is that separately trained modality adapters can be merged into a competent multimodal system by a brief alignment stage. On its own, the speech projector trained with ASR, ST, and SQA data could transcribe and translate but essentially could not perform spoken question answering; the text-only LoRA could handle QA on text but had no speech input. After the two are loaded together and instruction-tuned for 1K steps on mixed speech and text batches, the resulting model answers spoken questions at close to the text-only topline according to the authors' LLM-as-a-judge evaluation, improves ASR by about 1-2 WER points and translation BLEU over projector-only models, and handles all three tasks from the same set of weights. The authors interpret this as evidence that text and speech adaptation can be optimized independently and then aligned with a small amount of joint multimodal data.

Load-bearing premise

The spoken-QA gains are measured with an LLM-as-a-judge accuracy score that the paper's own Table 8 shows can diverge sharply from the official BERT-score metric for the same outputs, so if that judge is not a faithful measure, the central SQA claim is not established.

Editorial extensions

If this is right

  • A 1,000-step joint tuning stage is enough to fuse separately pretrained modality adapters; the paper's ablations show 2K steps mainly saturate or hurt, so the merge is cheap.
  • Adding the text modality during fusion improves speech translation most in the low-resource direction (Italian), suggesting textual supervision acts as a helpful regularizer when speech data are scarce.
  • Regenerating extractive QA answers as fluent sentences before training sharply improves spoken QA scores, at a small ASR cost, and is the variant chosen for the submitted system.
  • The speech projector alone, even when trained with SQA data, fails to produce SQA answers; the reasoning-style task only becomes learnable once the text LoRA and the projector are aligned in the C stage.

Reading between the lines

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

  • Editorial extension: the same parallel-then-align recipe may transfer to other modalities; a vision or audio adapter and a text-specialized LoRA could be pretrained independently and fused with a short joint stage, provided the alignment data is representative.
  • Editorial extension: the paper's own comparison between its LLM judge and the organizers' BERT-score protocol indicates the near-topline spoken-QA result is not yet settled; before using this model as a QA system, one should re-score outputs with the official metric.
  • Editorial extension: the fluent-answer regeneration step was motivated by domain shift from extractive answers, but the same idea could be tested as a general data augmentation for any extractive QA dataset feeding an LLM, since it converts exact spans into natural sentence-level supervision.
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

3 major / 2 minor

Summary. The paper describes NAVER LABS Europe's constrained submission to the IWSLT 2025 instruction-following speech processing track. The proposed pipeline trains a speech projector (A) on ASR/ST or ASR/ST/SQA data and text LoRA adapters (B) on MT/QA data in parallel, then merges them via a short joint instruction-tuning stage (C) of 1K steps. Results are reported for ASR, ST, and SQA on the ACL 60-60 and SpokenSQuAD test sets, with ablations on the adaptation stage, synthetic data, and task ratios. The central structural claim is that the parallel-training-plus-short-alignment recipe works: the multimodal models outperform the projector-only models on ASR and ST and make the speech model capable of SQA, which the projector alone could not do.

Significance. If confirmed, the A+B+C recipe is a practically useful and inexpensive way to add speech capabilities to a frozen instruction-tuned LLM without training a full multimodal model from scratch. The paper's strengths include clear descriptions of the data pipeline, detailed hyperparameter reporting, and a valuable control in Table 10 comparing a randomly initialized LoRA with the pretrained LoRA during the merging step. The paper is also transparent about the discrepancy between its LLM-as-a-judge SQA scores and the organizers' BERT-score results. However, the SQA conclusions are carried entirely by the LLM-as-judge metric, which the paper itself shows diverges sharply from the official protocol, and one explicit comparative claim is contradicted by the paper's own table. These issues need to be resolved before the central claims can be accepted.

major comments (3)
  1. [Section 5.2, Final Discussion and Table 2] The sentence in the Final Discussion that the models 'outperform SeamlessM4T-v2-large across all language pairs and metrics' is directly contradicted by Table 2. For en-it BLEU, SeamlessM4T-v2-large scores 43.54 while A.1+B (fluent) scores 40.48; for ASR WER, SeamlessM4T-v2-large scores 17.6 while the best multimodal models score 18.6. The claim should be corrected to identify the specific language pairs and metrics where the submission is better, and the exceptions should be acknowledged.
  2. [Appendix C.2.1, Tables 2 and 8] The SQA results in Table 2 are not supported by the paper's own Table 8. For the submitted model A.1+B (fluent) on valid en-de SQA, Table 2 reports 85.2% LLM-as-judge accuracy, while Table 8 reports a BERT score of 0.532 using the organizers' protocol. Table 8 also shows near-0.99 BERT scores on invalid questions, so the gap is not a uniform offset between metrics. Since the claims that multimodal training enables SQA and approaches the text-only topline are based entirely on the LLM-as-judge numbers, the SQA component of the central claim is not currently established. Please validate the LLM-as-judge metric against the reference-based metric, report both sets of numbers in the main table, or substantially soften the SQA claims.
  3. [Appendix C.2, 'Task Ratios' and Table 11] Table 11 reports ASR/ST/SQA task-ratio sweeps, and the text states that the ratio was selected as the one with the best average WER and BLEU over 'both ACL 60-60 dev and test set.' Because Table 2 reports results on the ACL 60-60 eval set, the reported ASR/ST numbers for the final configuration are affected by test-set selection. This makes the gains of the chosen task ratio optimistic and weakens the ablation's validity. Please rerun the sweep using only the dev set, or explicitly state that the final test-set numbers include selection bias.
minor comments (2)
  1. [Appendix C.2, 'Inclusion of Synthetic Textual Data'] The paragraph says 'Table 11 presents the results of our investigation into the inclusion of potentially noisy synthetic textual data,' but the synthetic-data rows (No synthetic data, Only synthetic data) appear in Table 10, not Table 11. The cross-reference should be corrected.
  2. [Table 1] The entry '34, 003†' contains an extra space inside the number; this is a formatting typo.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper is an empirical system description whose central claims are supported by benchmark comparisons against external baselines, not by a derivation from its own assumptions.

full rationale

This is an empirical system paper with no mathematical derivation targeted at a predicted quantity. The pipeline is: train a speech projector (A) on speech-to-text data, train text LoRA adapters (B) on text-to-text data, then jointly adapt both for 1K steps (C). The paper's central claim, that this parallel training followed by short joint instruction tuning works, is supported by measured results on external benchmarks (ACL 60-60 for ASR/ST, SpokenSQuAD for SQA) and by comparison with external baselines (SeamlessM4T-v2-large, Llama-3.1-8B-Instruct). SeamlessM4T-v2-large is used both to synthesize training data and as an evaluation baseline, but this is not circular: the baseline appears only in evaluation comparisons and is not embedded in the training objective. No fitted parameter is renamed as a prediction, no self-defined quantity is shown to equal another by construction, and no uniqueness theorem or ansatz is imported from the authors' prior work to force a choice. The paper contains self-citations, notably the bergen library [24] used for the LLM-as-judge evaluation, but this is tooling support, not load-bearing evidence for the central claim. The acknowledged discrepancy between the authors' LLM-judge SQA metric and the organizers' BERT-score evaluation (Appendix C.2.1, Tables 2 and 8) is a measurement-validity concern about the SQA results, not a circularity: the metric is not derived from the model's training objective or from the fitted parameters. Accordingly, no circular step can be exhibited, and the correct finding is no significant circularity.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central claim rests on a recipe of hyperparameters and frozen backbones rather than on free mathematical constants. The main new component is the speech projector architecture, but it is a standard Transformer encoder, not an invented entity.

free parameters (5)
  • Task sampling ratios for multimodal training = ASR 0.2, ST 0.4, SQA 0.4
    Selected by searching over five configurations and picking the best average WER/BLEU on both ACL 60-60 dev and test set (Table 11). This is test-set tuning.
  • Multimodal adaptation steps = 1,000
    Chosen from preliminary experiments comparing 100 and 1K steps; 1K gave the best gains (Section 5.1 C).
  • ST language sampling ratios (A.1) = en-de 0.3, en-zh 0.4, en-it 0.3
    Set proportionally to dataset size (Section 5.1 A.1).
  • COMET filter threshold for multilingual SQA = 0.85
    Used to filter synthetic translated questions and answers; threshold chosen without ablation (Section 2.1).
  • LoRA rank and alpha = rank=8, alpha=16
    Standard defaults from the LoRA paper, not tuned (Section 3).
assumptions (4)
  • domain assumption SeamlessM4T-v2-large encoder features suffice for a lightweight projector to map to Llama embeddings
    The pipeline freezes the encoder and trains a 4-layer transformer projector; the paper provides no error analysis on information loss from averaging every 3 frames. Invoked in Section 3.
  • domain assumption Llama-3.1-8B-Instruct can perform ASR, ST, and SQA given projected speech embeddings
    The projector and LoRA are trained on top of a frozen LLM; success depends on the LLM's latent capabilities. Invoked in Section 3.
  • domain assumption Synthetic data from SeamlessM4T-v2-large and Llama-3.1-8B-Instruct is faithful and useful for training
    A large share of training data is synthetic (Table 1); only COMET filtering (threshold 0.85) for translations and language ID for regenerated answers is used as quality control. Invoked in Sections 2.1 and Appendix A.2.
  • domain assumption LLM-as-a-judge scores reflect true SQA correctness
    Primary SQA metric is LLM-as-a-judge (Section 4), but the authors later find large discrepancies with the organizers' BERT-score protocol (Tables 2 vs 8, Appendix C.2.1).

how reviews work

0 comments
Cite this review

Pith. "Pith review of NAVER LABS Europe Submission to the Instruction-following Track." pith.science (2026). https://pith.science/paper/MCTK5QOX

@misc{pith2026250601808,
  author       = {Pith},
  title        = {Pith review of: NAVER LABS Europe Submission to the Instruction-following Track},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MCTK5QOX}},
  note         = {Machine review of arXiv:2506.01808}
}
read the original abstract

In this paper we describe NAVER LABS Europe submission to the instruction-following speech processing short track at IWSLT 2025. We participate in the constrained settings, developing systems that can simultaneously perform ASR, ST, and SQA tasks from English speech input into the following target languages: Chinese, Italian, and German. Our solution leverages two pretrained modules: (1) a speech-to-LLM embedding projector trained using representations from the SeamlessM4T-v2-large speech encoder; and (2) LoRA adapters trained on text data on top of a Llama-3.1-8B-Instruct. These modules are jointly loaded and further instruction-tuned for 1K steps on multilingual and multimodal data to form our final system submitted for evaluation.

Figures

Figures reproduced from arXiv: 2506.01808 by the authors.

Figure 1
Figure 1. Our training pipeline. A speech projector (A) and text LoRA adapters (B) are trained in parallel using [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

35 extracted references · 15 canonical work pages

  1. [1]

    McCrae, Salima Mdhaffar, Yasmin Moslem, Kenton Murray, Satoshi Nakamura, Matteo Negri, Jan Niehues, Atul Kr

    Idris Abdulmumin, Victor Agostinelli, Tanel Alumäe, Antonios Anastasopoulos, Ashwin, Luisa Bentivogli, Ondřej Bojar, Claudia Borg, Fethi Bougares, Roldano Cattoni, Mauro Cettolo, Lizhong Chen, William Chen, Raj Dabre, Yannick Estève, Marcello Federico, Marco Gaido, Dávid Javorský, Marek Kasztelnik, Tsz Kin Lam, Danni Liu, Evgeny Matusov, Chandresh Kumar M...

    work page 2025

  2. [2]

    Gpt-4 technical report.arXiv preprint arXiv:2303.08774, 2023

    Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, et al. Gpt-4 technical report.arXiv preprint arXiv:2303.08774, 2023. 1

    arXiv 2023

  3. [3]

    Duarte Miguel Alves, José Pombal, Nuno M Guerreiro, PedroHenriqueMartins, JoãoAlves, AminFarajian, Ben Peters, Ricardo Rei, Patrick Fernandes, Sweta Agrawal, Pierre Colombo, José G. C. de Souza, and Andre Mar- tins. Tower: Anopenmultilinguallargelanguagemodel for translation-related tasks. InFirst Conference on Lan- guage Modeling, 2024. 1

    work page 2024

  4. [4]

    From tower to spire: Adding the speech modality to a text-only llm.arXiv preprint arXiv:2503.10620, 2025

    Kshitij Ambilduke, Ben Peters, Sonal Sannigrahi, Anil Keshwani, Tsz Kin Lam, Bruno Martins, Marcely Zanon Boito, and André FT Martins. From tower to spire: Adding the speech modality to a text-only llm.arXiv preprint arXiv:2503.10620, 2025. 1

    arXiv 2025

  5. [5]

    Seamlessm4t: Massively multilingual & multimodal machine translation

    Loïc Barrault, Yu-An Chung, Mariano Cora Meglioli, David Dale, NingDong, Paul-Ambroise Duquenne, Hady Elsahar, Hongyu Gong, Kevin Heffernan, John Hoff- man, et al. Seamlessm4t: Massively multilingual & multimodal machine translation. arXiv preprint arXiv:2308.11596, 2023. 1

    arXiv 2023

  6. [6]

    Moshi: a speech-text foun- dation model for real-time dialogue

    Alexandre Défossez, Laurent Mazaré, Manu Orsini, Amélie Royer, Patrick Pérez, Hervé Jégou, Edouard Grave, and Neil Zeghidour. Moshi: a speech-text foun- dation model for real-time dialogue. arXiv preprint arXiv:2410.00037, 2024. 1

    arXiv

  7. [7]

    Palm- e: An embodied multimodal language model.arXiv preprint arXiv:2303.03378, 2023

    Danny Driess, Fei Xia, Mehdi SM Sajjadi, Corey Lynch, Aakanksha Chowdhery, Brian Ichter, Ayzaan Wahid, JonathanTompson, QuanVuong, TianheYu, etal. Palm- e: An embodied multimodal language model.arXiv preprint arXiv:2303.03378, 2023. 1

    arXiv 2023

  8. [8]

    The llama 3 herd of models

    Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, et al. The llama 3 herd of models. arXiv preprint arXiv:2407.21783, 2024. 1

    arXiv 2024

Show all 35 references
  1. [9]

    LoRA: Low-rank adaptation of large language 7 NAVER LABS Europe Submission to the Instruction-following Track models

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen- Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. LoRA: Low-rank adaptation of large language 7 NAVER LABS Europe Submission to the Instruction-following Track models. InInternational Conference on Learning Repre- se...

    2022

  2. [10]

    Wavllm: Towards ro- bust and adaptive speech large language model.arXiv preprint arXiv:2404.00656, 2024

    Shujie Hu, Long Zhou, Shujie Liu, Sanyuan Chen, Ling- wei Meng, Hongkun Hao, Jing Pan, Xunying Liu, Jinyu Li, Sunit Sivasankaran, et al. Wavllm: Towards ro- bust and adaptive speech large language model.arXiv preprint arXiv:2404.00656, 2024. 1

    2024 arXiv

  3. [11]

    Audiogpt: Understand- ing and generating speech, music, sound, and talking head

    Rongjie Huang, Mingze Li, Dongchao Yang, Jiatong Shi, Xuankai Chang, Zhenhui Ye, Yuning Wu, Zhiqing Hong, Jiawei Huang, Jinglin Liu, et al. Audiogpt: Understand- ing and generating speech, music, sound, and talking head. InProceedings of the AAAI Conference on Artificial Intel...

    2024

  4. [12]

    Iranzo-Sánchez, J

    J. Iranzo-Sánchez, J. A. Silvestre-Cerdà, J. Jorge, N. Roselló, A. Giménez, A. Sanchis, J. Civera, and A. Juan. Europarl-st: A multilingual corpus for speech transla- tion of parliamentary debates. InICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Sig...

    2020

  5. [13]

    Mixtral of experts.arXiv preprint arXiv:2401.04088, 2024

    Albert Q Jiang, Alexandre Sablayrolles, Antoine Roux, Arthur Mensch, Blanche Savary, Chris Bamford, De- vendra Singh Chaplot, Diego de las Casas, Emma Bou Hanna, Florian Bressand, et al. Mixtral of experts.arXiv preprint arXiv:2401.04088, 2024. 1

    2024 arXiv

  6. [14]

    What matters when building vision- language models?, 2024

    Hugo Laurençon, Léo Tronchon, Matthieu Cord, and Victor Sanh. What matters when building vision- language models?, 2024. arXiv:2405.02246. 1

    2024 arXiv

  7. [15]

    Spoken squad: A study of mitigating the impact ofspeechrecognitionerrorsonlisteningcomprehension

    Chia-Hsuan Lee, Szu-Lin Wu, Chi-Liang Liu, and Hung- yi Lee. Spoken squad: A study of mitigating the impact ofspeechrecognitionerrorsonlisteningcomprehension. Proc. Interspeech 2018, pages 3459–3463, 2018. 2

    2018

  8. [16]

    ROUGE: A package for automatic eval- uation of summaries

    Chin-Yew Lin. ROUGE: A package for automatic eval- uation of summaries. InText Summarization Branches Out, pages 74–81, Barcelona, Spain, 2004. Association for Computational Linguistics. 5

    2004

  9. [17]

    Visual Instruction Tuning (LLaVA), 2023

    Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. Visual Instruction Tuning (LLaVA), 2023. arXiv:2304.08485 [cs]. 1

    2023 arXiv

  10. [18]

    Guerreiro, Ricardo Rei, Duarte M

    Pedro Henrique Martins, Patrick Fernandes, João Alves, Nuno M. Guerreiro, Ricardo Rei, Duarte M. Alves, José Pombal, Amin Farajian, Manuel Faysse, Mateusz Kli- maszewski, Pierre Colombo, Barry Haddow, José G. C. de Souza, Alexandra Birch, and André F. T. Martins. Eu- rollm: Mu...

    2024

  11. [19]

    Eurollm: Multilingual language mod- els for europe.arXiv preprint arXiv:2409.16235, 2024

    Pedro Henrique Martins, Patrick Fernandes, João Alves, Nuno M Guerreiro, Ricardo Rei, Duarte M Alves, José Pombal, Amin Farajian, Manuel Faysse, Mateusz Kli- maszewski, et al. Eurollm: Multilingual language mod- els for europe.arXiv preprint arXiv:2409.16235, 2024. 1

    2024 arXiv

  12. [20]

    Spirit-lm: Interleaved spoken and written language model.Transactions of the Association for Computational Linguistics, 13:30–52, 2025

    Tu Anh Nguyen, Benjamin Muller, Bokai Yu, Marta R Costa-Jussa, Maha Elbayad, Sravya Popuri, Christophe Ropers, Paul-Ambroise Duquenne, Robin Algayres, Rus- lan Mavlyutov, et al. Spirit-lm: Interleaved spoken and written language model.Transactions of the Association for Comput...

    2025

  13. [21]

    Breaking language barriers in visual language models via multilingual textual regularization

    Iñigo Pikabea, Iñaki Lacunza, Oriol Pareras, Carlos Es- colano, Aitor Gonzalez-Agirre, Javier Hernando, and Marta Villegas. Breaking language barriers in visual language models via multilingual textual regularization. arXiv preprint arXiv:2503.22577, 2025. 4

    2025 arXiv

  14. [22]

    A call for clarity in reporting BLEU scores

    Matt Post. A call for clarity in reporting BLEU scores. In Proceedings of the Third Conference on Machine Transla- tion: ResearchPapers,pages186–191,Belgium,Brussels,

  15. [23]

    Scaling speech technology to 1,000+ languages.Jour- nal of Machine Learning Research, 25(97):1–52, 2024

    Vineel Pratap, Andros Tjandra, Bowen Shi, Paden Tomasello, Arun Babu, Sayani Kundu, Ali Elkahky, Zhao- heng Ni, Apoorv Vyas, Maryam Fazel-Zarandi, et al. Scaling speech technology to 1,000+ languages.Jour- nal of Machine Learning Research, 25(97):1–52, 2024. 4

    2024

  16. [24]

    BERGEN: A benchmarking library for retrieval-augmented generation

    David Rau, Hervé Déjean, Nadezhda Chirkova, Thibault Formal, Shuai Wang, Stéphane Clinchant, and Vas- silina Nikoulina. BERGEN: A benchmarking library for retrieval-augmented generation. InFindings of the Association for Computational Linguistics: EMNLP 2024, pages 7640–7663, ...

    2024

  17. [25]

    Ricardo Rei, José G. C. de Souza, Duarte Alves, Chrysoula Zerva, Ana C Farinha, Taisiya Glushkova, Alon Lavie, Luisa Coheur, and André F. T. Martins. COMET-22: Unbabel-IST 2022 submission for the met- rics shared task. InProceedings of the Seventh Conference on Machine Transla...

    2022

  18. [26]

    Audiopalm: A large language model that can speak and listen.arXiv preprint arXiv:2306.12925, 2023

    Paul K Rubenstein, Chulayuth Asawaroengchai, Duc Dung Nguyen, Ankur Bapna, Zalán Borsos, Félix de Chaumont Quitry, Peter Chen, Dalia El Badawy, Wei Han, Eugene Kharitonov, et al. Audiopalm: A large language model that can speak and listen.arXiv preprint arXiv:2306.12925, 2023. 1

    2023 arXiv

  19. [27]

    Evaluating multilingual speech translation under realistic condi- tions with resegmentation and terminology

    Elizabeth Salesky, Kareem Darwish, Mohamed Al- Badrashiny, Mona Diab, and Jan Niehues. Evaluating multilingual speech translation under realistic condi- tions with resegmentation and terminology. InProceed- ings of the 20th International Conference on Spoken Lan- guage Transla...

    2023

  20. [28]

    Salmonn: Towardsgenerichearingabilitiesforlargelan- guage models.arXiv preprint arXiv:2310.13289, 2023

    Changli Tang, Wenyi Yu, Guangzhi Sun, Xianzhao Chen, Tian Tan, Wei Li, Lu Lu, Zejun Ma, and Chao Zhang. Salmonn: Towardsgenerichearingabilitiesforlargelan- guage models.arXiv preprint arXiv:2310.13289, 2023. 1

    2023 arXiv

  21. [29]

    Gemini: a family of highly capable multimodal models

    Gemini Team, Rohan Anil, Sebastian Borgeaud, Jean- Baptiste Alayrac, Jiahui Yu, Radu Soricut, Johan Schalk- wyk, Andrew M Dai, Anja Hauth, Katie Millican, et al. Gemini: a family of highly capable multimodal models. arXiv preprint arXiv:2312.11805, 2023. 1

    2023 arXiv

  22. [30]

    Gemma Team, Aishwarya Kamath, Johan Ferret, Shreya Pathak, Nino Vieillard, Ramona Merhej, Sarah Per- rin, Tatiana Matejovicova, Alexandre Ramé, Morgane Rivière, Louis Rouillard, Thomas Mesnard, Geoffrey 8 NAVER LABS Europe Submission to the Instruction-following Track Cideron,...

    2025

  23. [31]

    torchtune: Py- torch’s finetuning library, 2024

    torchtune maintainers and contributors. torchtune: Py- torch’s finetuning library, 2024. 4

    2024

  24. [32]

    Llama 2: Open foundation and fine-tuned chat models, 2023

    Hugo Touvron, Louis Martin, Kevin Stone, Peter Al- bert, Amjad Almahairi, Yasmine Babaei, Nikolay Bash- lykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, Dan Bikel, Lukas Blecher, Cristian Canton Ferrer, Moya Chen, Guillem Cucurull, David Esiobu, Jude Fernan- des, Jeremy...

    2023

  25. [33]

    Covost 2: A massively multilingual speech-to-text translation corpus, 2020

    Changhan Wang, Anne Wu, and Juan Pino. Covost 2: A massively multilingual speech-to-text translation corpus, 2020. 2

    2020

  26. [34]

    1”, instead of “0

    An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, et al. Qwen2. 5 technical report. arXiv preprint arXiv:2412.15115, 2024. 1 9 NAVER LABS Europe Submission to the Instruction-following Track A. Data Prepro...

    2024 arXiv

  27. [2018]

    Association for Computational Linguistics. 4

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.