Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

Probing Representations Learned by Multimodal Recurrent and Transformer Models

T0 review · 3 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read Sentence representations learn more from translation and image grounding than from language modeling alone.

desk verdict Useful meta-study showing grounded training beats LM for sentence representations on moderate data, but the RNN-vs-Transformer semantic claim is built on unmatched models and single runs. read the letter →

arxiv 1908.11125 v1 pith:F4WJ3LTG submitted 2019-08-29 cs.CL

Jindřich Libovický , Pranava Madhyastha This is my paper · ORCID
classification cs.CL
keywords sentencerepresentationsmultimodalmachinetranslationlanguagemodelingsemantictextualsimilarityimageretrievalcanonicalcorrelationanalysisrecurrentneuralnetworksTransformer
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 asks what makes a sentence representation semantically useful: predicting the next word, predicting an image's visual features, or producing a sentence in another language. Across models trained on the same moderate-sized data, the paper finds that translation and image grounding give a much stronger training signal than language modeling. It further finds that although Transformer models translate better, recurrent (RNN) encoders produce representations that score higher on semantic textual similarity and image retrieval. The authors interpret this as evidence that the architecture of the encoder shapes what information the representation preserves, not just how well the model performs its training task.

What carries the argument

The machinery is a controlled comparison of four training objectives—language modeling, image-feature prediction (Imaginet), text-only machine translation, and multimodal machine translation—each implemented with an RNN and a Transformer architecture. The representations are evaluated without fine-tuning by two probes: canonical correlation analysis (CCA) is fitted on sentence–image pairs to project both spaces and measure retrieval recall at 10, and cosine distance between sentence vectors is measured against human ratings in the Semantic Textual Similarity benchmark, reported as Spearman correlation. A third tool, distance correlation, quantifies how much two learned representation spaces depend on each other, capturing both linear and nonlinear association. This setup isolates the training signal's contribution to representation quality while holding the downstream probes fixed.

What would settle it

Train an RNN and a Transformer encoder with matched parameter counts, layer depths, and multiple random seeds on the same translation objective, then probe both on the same STS and image-retrieval tasks; if the Transformer matches or exceeds the RNN's semantic scores, the paper's architectural claim is refuted.

Watch

Extended reading notes

Core claim

The paper's central discovery is a dissociation between task performance and representation quality. On the Multi30k benchmark and its extensions, models that predict a target-language sentence or an image representation learn sentence encoders whose outputs correlate far better with human semantic similarity judgments than encoders trained only on language modeling, and the effect grows with access to richer training signals. At the same time, the highest translation quality is reached by Transformer models, yet the representations of RNN-based encoders are the ones that score better on semantic textual similarity and on retrieval of the image described by a sentence. The paper also finds that multimodal translation models that explicitly condition the decoder on image features produce encoders that largely ignore visual information, and that translation quality, image-retrieval ability, and semantic similarity are all positively correlated.

Load-bearing premise

The load-bearing assumption is that the RNN and Transformer models are comparable enough that the RNN's better semantic scores can be attributed to the architecture itself, even though the two use different depths, widths, and training dynamics.

Editorial extensions

If this is right

  • If grounding is the stronger signal, then for low-resource languages or domains where parallel data are scarce, image-caption pairs or any cross-modal alignment should be prioritized over additional monolingual text for representation learning.
  • RNN encoders may remain preferable for semantic search, paraphrase detection, and retrieval systems even where Transformers give better generation quality.
  • The positive correlation between BLEU and STS means that improving translation quality should also improve semantic representation quality, so machine translation progress is directly reusable for semantic tasks.
  • Multimodal models that condition the decoder on images teach the encoder to offload visual information, so encoder representations should be evaluated separately from the full model's output.

Reading between the lines

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

  • Editorial extension: the same training-signal comparison could be run with the language modeling objective scaled to the same parameter count as the translation models, which would test whether the apparent RNN advantage is architectural or a capacity artifact.
  • Editorial extension: the probing methodology could be transferred to large pretrained models, fine-tuning them on translation or image-prediction objectives and probing on STS to see whether grounding still helps once massive monolingual data are already available.
  • Editorial extension: the paper's result that sentence representations cluster by task rather than architecture suggests that the training objective, not the sequence model, is the dominant determinant of representational geometry, which predicts that any sufficiently expressive architecture will produce similar representation clusters under the same objective.
  • Editorial extension: a practical consequence the authors do not spell out is that distilling Transformer translation models into RNN student encoders could yield semantic embeddings that combine the translation quality of the teacher with the representational properties of the student.
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 / 6 minor

Summary. This paper presents a meta-study of sentence representations extracted from recurrent (RNN) and Transformer models trained on four objectives: language modeling, image-feature prediction (Imaginet), textual machine translation, and multimodal machine translation. Representations are evaluated without fine-tuning on image retrieval via canonical correlation analysis and on semantic textual similarity (STS) via cosine similarity, and representation similarity is measured with distance correlation. The main empirical claims are that (i) on moderate-sized datasets, supervision from a target language or visual modality provides a stronger training signal than language modeling, and (ii) while Transformers achieve higher BLEU scores, RNN-based models yield representations that perform better on semantic relevance tasks, especially STS. The paper also reports that multimodal encoders tend to ignore image information when the decoder has direct access to it, and that MT model BLEU correlates strongly with STS performance. The experiments use the Multi30k, Flickr30k, and COCO datasets and compare against ELMo and BERT as large-scale pretrained baselines.

Significance. If the main claims were established, the paper would be a useful empirical contribution to the discussion of what makes sentence representations semantically useful, and it provides a reasonably clean evaluation protocol by keeping representations fixed and using external benchmarks (SemEval 2016 STS, CCA-based image retrieval). The breadth of models and training signals is a strength, and the distance-correlation analysis adds a non-linear dependence view that is rarely included in such studies. However, the headline architectural claim—that RNN representations are semantically better than Transformer representations—is currently supported only by confounded comparisons: the architectures differ in depth, capacity, and representation dimensionality, and all results come from single runs with no variance estimates or significance tests. The first claim (grounding beats language modeling) is also partly confounded by bidirectionality. The paper is therefore a useful preliminary study whose central causal interpretation needs additional experimental support.

major comments (3)
  1. [§3.1 and Table 1] The central RNN-vs-Transformer comparison is confounded by capacity, depth, and representation dimensionality. The RNN encoder is a single bidirectional GRU with 1000 units and 500-dimensional embeddings, so the mean-pooled sentence representation is 2000-dimensional, whereas the Transformer encoder is 6-layer, 512-dimensional, with 8 heads, giving a 512-dimensional mean-pooled representation. Cosine STS and CCA-based image retrieval are both sensitive to dimensionality and optimization, so the observed advantage of RNN models on STS (e.g., hierarchical attention RNN MMT .553 vs. parallel Transformer MMT .398) could stem from these differences rather than from an architectural inductive bias. The paper needs a matched-control comparison—equal parameter counts or at least equal representation dimensionality, with evaluation done on projections to a common dimensionality—before the abstract's architecture-level claim is supportable.
  2. [§3.1 and Table 1 (LM rows)] The claim that grounding provides a stronger training signal than language modeling is partly confounded by directionality. The RNN language model is described as a single GRU layer and the Transformer LM is presumably a standard left-to-right language model, while the Imaginet, MT, and multimodal MT encoders are either bidirectional RNNs or full Transformer encoders with bidirectional self-attention. The comparison 'grounded model beats LM' therefore conflates the training signal with access to bidirectional context. A control experiment with a bidirectional LM, or with unidirectional grounded encoders, is needed to attribute the improvement to grounding rather than to context directionality.
  3. [§4 and Table 1] The abstract and Section 4 use the word 'significantly' (e.g., 'representations from the recurrent neural network based models perform significantly better'), but all models are trained once with no seeds, error bars, or significance tests. Some STS differences are large (RNN MMT hierarchical .553 vs. Transformer parallel .398), but others are small (Multi30k LM RNN .267 vs. Transformer .256), and image retrieval differences are sometimes reversed (Imagination models). Without multiple seeds or confidence intervals, the 'significant' language is unsupported. Reporting means over at least three seeds, with standard deviations, and ideally a paired significance test across evaluation conditions, would make the main comparisons credible.
minor comments (6)
  1. [Table 1] The first column header says 'BLUE score'; this should be 'BLEU score'.
  2. [§4 (Figure 1)] The paragraph beginning 'that RNN models obtain semantically richer representations' and the Figure 1 caption appear twice in the text; this appears to be a copy-paste error and should be removed.
  3. [§5] 'We conducted a set of controlled and thorough experiments to asses the representational qualities'—'asses' should be 'assess'.
  4. [§4] The sentence 'but the other way round with Transformer based Imagination models that are explicitly trained to predict the image representation perform better than their RNN counterparts' is grammatically unclear and should be rewritten for readability.
  5. [§3.1] The manuscript says 'All models trained with Neural Monkey'; this should be 'All models were trained using Neural Monkey.'.
  6. [§3.1] It is not explicitly stated whether the Transformer language model is trained with a causal/left-to-right mask; given the confound discussed above, this should be stated explicitly.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the empirical comparisons rest on fixed representations and external benchmarks, with only minor non-load-bearing self-citations.

full rationale

The paper is an empirical meta-study rather than a derivation chain. Sentence representations are extracted from models trained on standard objectives (language modeling, image-feature prediction, textual and multimodal MT), and the evaluation uses external benchmarks: SemEval 2016 STS and CCA-based image retrieval on the Multi30k test split. The only learned projection, CCA, is explicitly fit on the training portion and applied to held-out test pairs, and the paper states that the correlation is not backpropagated into the representations ('Note that we do not backpropagate the correlation to the network and keep the representation fixed because our goal is not training towards optimal cross-modal representation but only to asses the (already trained) sentence representation'). The central claims comparing RNN and Transformer representations therefore do not reduce by construction to fitted parameters or to the evaluation metrics themselves. The self-citations present (Helcl and Libovický 2017 for Neural Monkey; Libovický and Helcl 2017 and Libovický et al. 2018 for multi-source attention strategies) are methodological citations to model architectures and tools, not load-bearing evidence for the empirical ranking; no uniqueness theorem or ansatz is imported from these citations. The capacity mismatch between the single-layer GRU and the 6-layer Transformer noted in the reader's take is a legitimate internal-validity concern about whether architecture, capacity, or optimization drives the observed differences, but it is not a circularity: the RNN-versus-Transformer comparison is an empirical result, not something defined into existence by the paper's equations or by a self-citation chain. Consequently, no circular steps are identified, and the appropriate score is 0.

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

The paper introduces no mathematical derivation and no new entities. Its conclusions rest on standard evaluation assumptions and an uncontrolled capacity comparison between architectures.

assumptions (4)
  • domain assumption Mean-pooled encoder hidden states are a faithful sentence representation for downstream evaluation.
    Used to extract representations for all models in the evaluation.
  • domain assumption Cosine distance between sentence representations is a valid proxy for human semantic similarity.
    STS evaluation assumes this monotonic relationship.
  • domain assumption Linear CCA projections estimated on the Multi30k training split generalize to the test set for image retrieval.
    Image retrieval evaluation relies on this.
  • domain assumption The training objectives (LM, image prediction, MT, MMT) were optimized to comparable convergence under the Neural Monkey toolkit, making their representations comparable.
    Comparability of models across tasks assumes training was fair.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Probing Representations Learned by Multimodal Recurrent and Transformer Models." pith.science (2026). https://pith.science/paper/F4WJ3LTG

@misc{pith2026190811125,
  author       = {Pith},
  title        = {Pith review of: Probing Representations Learned by Multimodal Recurrent and Transformer Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/F4WJ3LTG}},
  note         = {Machine review of arXiv:1908.11125}
}
read the original abstract

Recent literature shows that large-scale language modeling provides excellent reusable sentence representations with both recurrent and self-attentive architectures. However, there has been less clarity on the commonalities and differences in the representational properties induced by the two architectures. It also has been shown that visual information serves as one of the means for grounding sentence representations. In this paper, we present a meta-study assessing the representational quality of models where the training signal is obtained from different modalities, in particular, language modeling, image features prediction, and both textual and multimodal machine translation. We evaluate textual and visual features of sentence representations obtained using predominant approaches on image retrieval and semantic textual similarity. Our experiments reveal that on moderate-sized datasets, a sentence counterpart in a target language or visual modality provides much stronger training signal for sentence representation than language modeling. Importantly, we observe that while the Transformer models achieve superior machine translation quality, representations from the recurrent neural network based models perform significantly better over tasks focused on semantic relevance.

Figures

Figures reproduced from arXiv: 1908.11125 by the authors.

Figure 1
Figure 1. Distance correlation of representations from if ltd dl Figure 1: Distance correlation of representations from pairs of selected models Figure 1: Distance correlation of representations from [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Intrinsic vs. Extrinsic Evaluation of Czech Sentence Embeddings: Semantic Relevance Doesn't Help with MT Evaluation

    cs.CL 2025-06 conditional novelty 6.0 of 10

    For English-to-Czech MT evaluation, embeddings that win intrinsic semantic similarity benchmarks (SimCSE) perform worst after fine-tuning, while intrinsically poor models (XLM-R, FERNET) perform best.

Reference graph

Works this paper leans on

41 extracted references · 17 canonical work pages · cited by 1 Pith paper

  1. [1]

    URL: " 'urlintro :=

    ENTRY address author booktitle chapter edition editor howpublished institution journal key month note number organization pages publisher school series title type volume year eprint doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRINGS urlintro eprinturl eprintpr...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Eneko Agirre, Carmen Banea, Daniel Cer, Mona Diab, Aitor Gonzalez-Agirre, Rada Mihalcea, German Rigau, and Janyce Wiebe. 2016. https://doi.org/10.18653/v1/s16-1081 Semeval-2016 task 1: Semantic textual similarity, monolingual and cross-lingual evaluation . In Proceedings of the 10th International Workshop on Semantic Evaluation (SemEval-2016), pages 497--...

    doi:10.18653/v1/s16-1081 2016

  4. [4]

    Eneko Agirre, Daniel Cer, Mona Diab, and Aitor Gonzalez-Agirre. 2012. https://doi.org/10.3115/v1/s12-1051 Semeval-2012 task 6: A pilot on semantic textual similarity . In *SEM 2012 : The First Joint Conference on Lexical and Computational Semantics -- Volume 1: Proceedings of the main conference and the shared task, and Volume 2: Proceedings of the Sixth ...

    work page doi:10.3115/v1/s12-1051 2012

  5. [5]

    Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Bengio. 2014. https://arxiv.org/abs/1409.0473 Neural machine translation by jointly learning to align and translate . CoRR, abs/1409.0473

    arXiv 2014

  6. [6]

    Ozan Caglayan, Walid Aransa, Adrien Bardet, Mercedes Garc\' i a-Mart\' i nez, Fethi Bougares, Lo\" i c Barrault, Marc Masana, Luis Herranz, and Joost van de Weijer. 2017. https://doi.org/10.18653/v1/w17-4746 Lium-cvc submissions for wmt17 multimodal translation task . In Proceedings of the Second Conference on Machine Translation, pages 432--439, Copenhag...

    work page doi:10.18653/v1/w17-4746 2017

  7. [7]

    Iacer Calixto and Qun Liu. 2017. https://doi.org/10.18653/v1/d17-1105 Incorporating global visual features into attention-based neural machine translation. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing ( EMNLP ) , pages 992--1003, Copenhagen, Denmark. Association for Computational Linguistics

    doi:10.18653/v1/d17-1105 2017

  8. [8]

    Kyunghyun Cho, Bart van Merrienboer, Dzmitry Bahdanau, and Yoshua Bengio. 2014. https://doi.org/10.3115/v1/w14-4012 On the properties of neural machine translation: Encoder--decoder approaches . In Proceedings of SSST-8, Eighth Workshop on Syntax, Semantics and Structure in Statistical Translation, pages 103--111, Doha, Qatar. Association for Computationa...

    doi:10.3115/v1/w14-4012 2014

Show all 41 references
  1. [9]

    Grzegorz Chrupa a, \' A kos K\' a d\' a r, and Afra Alishahi. 2015. https://doi.org/10.3115/v1/p15-2019 Learning language through pictures . In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on...

    2015 doi

  2. [10]

    Ond r ej C \' fka and Ond r ej Bojar. 2018. https://doi.org/10.18653/v1/P18-1126 Are BLEU and meaning representation in opposition? In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1362--1371, Melbourne, ...

    2018 doi

  3. [11]

    Alexis Conneau and Douwe Kiela. 2018. http://aclweb.org/anthology/L18-1269 SentEval : An evaluation toolkit for universal sentence representations . In Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC), pages 1699--1704, Miyazaki,...

    2018

  4. [12]

    Alexis Conneau, Douwe Kiela, Holger Schwenk, Lo\" i c Barrault, and Antoine Bordes. 2017. https://arxiv.org/abs/1705.02364 Supervised learning of universal sentence representations from natural language inference data . CoRR, 1705.02364

    2017 arXiv

  5. [13]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2018. https://arxiv.org/abs/1810.04805 Bert: Pre-training of deep bidirectional transformers for language understanding . CoRR, abs/1810.04805

    2018 arXiv

  6. [14]

    Desmond Elliott. 2018. https://doi.org/10.18653/v1/d18-1329 Adversarial evaluation of multimodal machine translation . In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing ( EMNLP ) , pages 2974--2978, Brussels, Belgium. Association for Com...

    2018 doi

  7. [15]

    Desmond Elliott, Stella Frank, Khalil Sima'an, and Lucia Specia. 2016. https://doi.org/10.18653/v1/W16-3210 Multi30k: Multilingual english-german image descriptions . In Proceedings of the 5th Workshop on Vision and Language, pages 70--74, Berlin, Germany. Association for Comp...

    2016 doi

  8. [16]

    Desmond Elliott and \' A kos K\' a d\' a r. 2017. https://doi.org/10.18653/v1/i17-1014 Imagination improves multimodal translation . In Proceedings of the Eighth International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 130--141, Taipei, Taiw...

    2017 doi

  9. [17]

    Orhan Firat and Kyunghyun Cho. 2016. Conditional gated recurrent unit with attention mechanism. https://github.com/nyu-dl/dl4mt-tutorial/blob/master/docs/cgru.pdf. Published online, version adbaeea

    2016

  10. [18]

    Daniela Gerz, Ivan Vuli\' c , Felix Hill, Roi Reichart, and Anna Korhonen. 2016. https://doi.org/10.18653/v1/d16-1235 Simverb-3500: A large-scale evaluation set of verb similarity . In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing ( EMN...

    2016 doi

  11. [19]

    Yunchao Gong, Liwei Wang, Micah Hodosh, Julia Hockenmaier, and Svetlana Lazebnik. 2014. https://doi.org/10.1007/978-3-319-10593-2_35 Improving image-sentence embeddings using large weakly annotated photo collections . In Computer Vision -- ECCV 2014, pages 529--545, Cham, Swit...

    2014 doi

  12. [20]

    David R Hardoon, Sandor Szedmak, and John Shawe-Taylor. 2004. https://doi.org/10.1162/0899766042321814 Canonical correlation analysis: An overview with application to learning methods . Neural Computation, 16(12):2639--2664

    2004 doi

  13. [21]

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. https://doi.org/10.1109/cvpr.2016.90 Deep residual learning for image recognition . In Conference on Computer Vision and Pattern Recognition ( CVPR ) , pages 770--778, Las Vegas, NV, USA. IEEE Computer Society

    2016 doi

  14. [22]

    Jind r ich Helcl and Jind r ich Libovick\' y . 2017 a . https://doi.org/10.18653/v1/w17-4749 CUNI system for the WMT17 multimodal translation task . In Proceedings of the Second Conference on Machine Translation, Volume 2: Shared Task Papers, pages 450--457, Copenhagen, Denmar...

    2017 doi

  15. [23]

    Jind r ich Helcl and Jind r ich Libovick\' y . 2017 b . https://doi.org/10.1515/pralin-2017-0001 Neural monkey: An open-source tool for sequence learning . The Prague Bulletin of Mathematical Linguistics, 107(1):5--17

    2017 doi

  16. [24]

    Jind r ich Helcl, Jind r ich Libovick\' y , and Du s an Vari s . 2018. https://doi.org/10.18653/v1/w18-6441 CUNI system for the WMT18 multimodal translation task . In Proceedings of the Third Conference on Machine Translation, pages 622--629, Brussels, Belgium. Association for...

    2018 doi

  17. [25]

    Felix Hill, Roi Reichart, and Anna Korhonen. 2015. https://doi.org/10.1162/coli_a_00237 Simlex-999: Evaluating semantic models with (genuine) similarity estimation . Computational Linguistics, 41(4):665--695

    2015 doi

  18. [26]

    Harold Hotelling. 1936. https://doi.org/10.2307/2333955 Relations between two sets of variates . Biometrika, 28(3/4):321--377

    1936 doi

  19. [27]

    Jeremy Howard and Sebastian Ruder. 2018. https://arxiv.org/abs/1801.06147 Fine-tuned language models for text classification . CoRR, abs/1801.06147

    2018 arXiv

  20. [28]

    Angeliki Lazaridou, Nghia The Pham, and Marco Baroni. 2015. https://doi.org/10.3115/v1/n15-1016 Combining language and vision with a multimodal skip-gram model . In Proceedings of the 2015 Conference of the North American Chapter of the Association for Computational Linguistic...

    2015 doi

  21. [29]

    Jind r ich Libovick\' y and Jind r ich Helcl. 2017. https://doi.org/10.18653/v1/p17-2031 Attention strategies for multi-source sequence-to-sequence learning . In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), p...

    2017 doi

  22. [30]

    Jind r ich Libovick\' y , Jind r ich Helcl, and David Mare c ek. 2018. https://doi.org/10.18653/v1/w18-64026 Input combination strategies for multi-source transformer decoder . In Proceedings of the Third Conference on Machine Translation, pages 253--260, Brussels, Belgium. As...

    2018 doi

  23. [31]

    Tsung-Yi Lin, Michael Maire, Serge J Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll\' a r, and Lawrence Zitnick. 2014. https://doi.org/10.1007/978-3-319-10602-1_48 Microsoft coco: Common objects in context . In Computer Vision -- ECCV 2014, pages 740--755, Cham,...

    2014 doi

  24. [32]

    Tom \'a s Mikolov, Martin Karafi \'a t, Luk \'a s Burget, Jan C ernock \`y , and Sanjeev Khudanpur. 2010. https://www.isca-speech.org/archive/interspeech_2010/i10_1045.html Recurrent neural network based language model . In Eleventh Annual Conference of the International Speec...

    2010

  25. [33]

    Matthew Peters, Mark Neumann, Mohit Iyyer, Matt Gardner, Christopher Clark, Kenton Lee, and Luke Zettlemoyer. 2018. https://doi.org/10.18653/v1/n18-1202 Deep contextualized word representations . In Proceedings of the 2018 Conference of the North American Chapter of the Associ...

    2018 doi

  26. [34]

    Bryan A Plummer, Liwei Wang, Chris M Cervantes, Juan C Caicedo, Julia Hockenmaier, and Svetlana Lazebnik. 2015. https://doi.org/10.1109/iccv.2015.303 Flickr30k entities: Collecting region-to-phrase correspondences for richer image-to-sentence models . In Proceedings of the 201...

    2015 doi

  27. [35]

    Nina Poerner, Benjamin Roth, and Hinrich Sch\" u tze. 2018. https://doi.org/10.18653/v1/w18-5437 Interpretable textual neuron representations for nlp . In Proceedings of the 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP, pages 325--327, Br...

    2018 doi

  28. [36]

    Naomi Saphra and Adam Lopez. 2018. https://doi.org/10.18653/v1/w18-5436 Language models learn pos first . In Proceedings of the 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP, pages 328--330, Brussels, Belgium. Association for Computational...

    2018 doi

  29. [37]

    Gábor J Székely, Maria L Rizzo, and Nail K Bakirov. 2007. https://doi.org/10.1214/009053607000000505 Measuring and testing dependence by correlation of distances . The Annals of Statistics, 35(6):2769--2794

    2007 doi

  30. [38]

    J\" o rg Tiedemann. 2012. http://www.lrec-conf.org/proceedings/lrec2012/pdf/463_Paper.pdf Parallel data, tools and interfaces in opus . In Proceedings of the 8th International Conference on Language Resources and Evaluation (LREC), pages 2214--2218, Istanbul, Turkey. European ...

    2012

  31. [39]

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, ukasz Kaiser, and Illia Polosukhin. 2017. https://papers.nips.cc/paper/7181-attention-is-all-you-need.pdf Attention is all you need . In Advances in Neural Information Processing Systems 30...

    2017

  32. [40]

    Alex Wang, Amanpreet Singh, Julian Michael, Felix Hill, Omer Levy, and Samuel Bowman. 2018. https://doi.org/10.18653/v1/w18-5446 GLUE : A multi-task benchmark and analysis platform for natural language understanding . In Proceedings of the 2018 EMNLP Workshop BlackboxNLP: Anal...

    2018 doi

  33. [41]

    Fei Yan and Krystian Mikolajczyk. 2015. https://doi.org/10.1109/cvpr.2015.7298966 Deep correlation for matching images and text . In Conference on Computer Vision and Pattern Recognition ( CVPR ) , pages 3441--3450, Boston, MA, USA. IEEE Computer Society

    2015

Pith tools

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