pith. sign in

arxiv: 2607.00171 · v1 · pith:I4JAJ2YZnew · submitted 2026-06-30 · 💻 cs.CL

ALEE: Any-Language Evaluation of Embeddings via English-Centric Minimal Pairs

Andrianos Michail , Stylianos Psychias , Michelle Wastl , Simon Clematide , Rico Sennrich , Juri Opitz This is my paper

Pith reviewed 2026-07-02 19:08 UTC · model grok-4.3

classification 💻 cs.CL
keywords text embeddingsevaluationcross-lingualminimal pairsAMRsemantic similaritylow-resource languagessubword tokenization
0
0 comments X

The pith

ALEE generates English AMR minimal pairs and their translations to diagnose embedding performance across 275+ languages.

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

The paper presents ALEE as a method that starts with Abstract Meaning Representations to produce English minimal pairs differing by controlled fine-grained semantic shifts, then pairs each with translations into target languages. This setup supports evaluation of any embedding model on any language for which English parallel text exists, moving beyond static or limited benchmarks. Large-scale tests on diverse models show that scores differ markedly by language, by text length, and by the type of linguistic change being tested. The differences align with how much training data each language received and with choices in subword tokenization.

Core claim

ALEE extends prior minimal-pair methods to the cross-lingual and paragraph setting by using AMR to create English pairs that isolate specific semantic distinctions and then translating those pairs. When applied to many embedding models across three parallel datasets covering more than 275 languages, the resulting scores expose consistent performance gaps that track language prevalence in training resources and subword tokenization practices, indicating that current embeddings still lack uniform cross-lingual semantic fidelity.

What carries the argument

AMR-generated English minimal pairs with controlled fine-grained semantic shifts, paired with translations into target languages for cross-lingual comparison.

If this is right

  • Targeted diagnostics become feasible for embedding models in any language supplied with English parallel data.
  • Scores vary substantially by language, by text length, and by linguistic phenomenon.
  • Gaps in cross-lingual semantic representation track language prevalence in training resources and subword tokenization.
  • The same framework supports evaluation at both sentence and paragraph levels across three large parallel datasets.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The observed correlations suggest that increasing training data volume for low-prevalence languages would narrow the measured gaps.
  • The method could be used to compare tokenization schemes by holding the underlying model fixed and varying only the vocabulary.
  • Extending ALEE to measure how well models handle phenomena that are rare in English but common in the target language would test an implicit boundary of the English-centric design.

Load-bearing premise

Translations of the AMR-generated English minimal pairs faithfully preserve the controlled fine-grained semantic shifts without introducing artifacts or losing the intended meaning distinctions in the target languages.

What would settle it

A direct comparison in which human raters judge semantic similarity on the translated pairs and those human ratings fail to align with the ordering produced by embedding cosine scores would show that the controlled distinctions are not preserved.

Figures

Figures reproduced from arXiv: 2607.00171 by Andrianos Michail, Juri Opitz, Michelle Wastl, Rico Sennrich, Simon Clematide, Stylianos Psychias.

Figure 1
Figure 1. Figure 1: Controlled cross-lingual polarity flip in [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Mean TACC per model broken down by semantic edit type, shown separately for ALEE-F200 (top), ALEE-MT61 (middle) and ALEE-BQ275 (bot￾tom). Models are sorted by their average across minimal pair types. their performance varies across the four semantic perturbation types. Setup. For each model and each transformation, we compute TACC averaged over languages, sepa￾rately for each ALEE dataset. Findings [PITH_… view at source ↗
Figure 3
Figure 3. Figure 3: ALEE-MT61, ALEE-BQ275 Macro-TACC across different text lengths. leads on ALEE-F200 and ALEE-BQ275, while LaBSE—a model trained for bitext mining with hard negatives—edges ahead on ALEE-MT61. At the same time, no model reaches perfect accuracy, indicating that all models fail on a non-trivial subset of minimal pairs. (2) Across all three datasets, po￾larity negation is consistently the easiest perturba￾tion… view at source ↗
Figure 4
Figure 4. Figure 4: Per-language Macro-TACC vs. CC-100 pretraining data size (log-scaled), averaged across all models, [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: ALEE-F200 Macro-TACC distribution by Common Crawl language prevalence. ALEE-MT61 and ALEE-BQ275 show similar patterns (Figs. 11 and 12). Findings [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: multilingual-gte-base. Per-language TACC vs. training data size for pretraining tokens (top) and post [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Per-language Macro-TACC vs. median XLM￾R subtoken count on ALEE-MT61, averaged across all models. Marker shapes indicate writing script. Figures 9 & 10 in the Appendix show ALEE-F200/BQ275 results. multilingual_e5_large_instruct sentence_transformers_LaBSE multilingual_e5_large BAAI_bge_m3 Alibaba_NLP_gte_multilingual_base multilingual_e5_base google_embeddinggemma_300m ibm_granite_embed_107m_multiL ibm_gr… view at source ↗
Figure 9
Figure 9. Figure 9: Per-language Macro-TACC vs. median XLM-R subtoken count on [PITH_FULL_IMAGE:figures/full_fig_p014_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Per-language Macro-TACC vs. median XLM-R subtoken count on [PITH_FULL_IMAGE:figures/full_fig_p014_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: ALEE-MT61 Macro-TACC distribution by Common Crawl language prevalence. Analogous to [PITH_FULL_IMAGE:figures/full_fig_p015_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: ALEE-BQ275 Macro-TACC distribution by Common Crawl language prevalence. Analogous to [PITH_FULL_IMAGE:figures/full_fig_p015_12.png] view at source ↗
read the original abstract

Text embeddings are standard for semantic similarity tasks, yet their evaluation remains an open challenge. Current benchmarks are static, cover only a limited set of languages, are often domain-specific, susceptible to overfitting, and poorly representative of low-resource languages. To address these limitations, we introduce ALEE, a framework that extends Sentence Smith (Li et al., 2025) to the cross-lingual and paragraph level. ALEE uses Abstract Meaning Representations (AMR) to generate English minimal pairs with controlled, fine-grained semantic shifts, which are paired with translations in target languages. This approach enables targeted diagnostics for models in any language with English parallel data. We conduct a large-scale empirical study across a diverse set of embedding models and 275+ languages spanning three parallel datasets. On ALEE, performance varies substantially across languages, text lengths, and linguistic phenomena, exposing persistent gaps in cross-lingual semantic representation that track language prevalence in training resources and subword tokenization. We release ALEE at https://github.com/Andrian0s/any-lang-embed-eval

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper introduces ALEE, a framework extending Sentence Smith to cross-lingual and paragraph-level evaluation. It generates controlled English minimal pairs via AMR, translates them to target languages, and uses the resulting pairs to diagnose embedding models across 275+ languages and three parallel datasets. The central empirical claim is that performance varies substantially by language, text length, and linguistic phenomenon, with gaps tracking training-data prevalence and subword tokenization.

Significance. If the translations faithfully preserve the intended fine-grained semantic contrasts, ALEE would supply a scalable, language-agnostic diagnostic that addresses the static, limited-coverage, and overfitting problems of existing embedding benchmarks while reaching low-resource languages.

major comments (2)
  1. [Abstract] Abstract: the claim that ALEE 'exposes persistent gaps in cross-lingual semantic representation' rests on the unverified assumption that translations of the AMR minimal pairs preserve the controlled semantic shifts without introducing artifacts or losing distinctions. No translation-quality controls, back-translation checks, or human fidelity ratings are described, which directly threatens the targeted-diagnostic interpretation, especially for low-resource languages where MT quality is lower.
  2. [Abstract] Abstract: the reported empirical variation across languages, lengths, and phenomena is presented without any mention of statistical significance testing, data-exclusion criteria, or sensitivity to implementation choices in the translation pipeline, leaving the load-bearing cross-lingual claims difficult to evaluate for robustness.
minor comments (1)
  1. [Abstract] The citation 'Sentence Smith (Li et al., 2025)' appears in the abstract; the reference list should confirm whether this is a published or forthcoming work and provide full bibliographic details.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for highlighting these important methodological concerns. We agree that explicit translation fidelity verification and statistical robustness checks would strengthen the paper's claims. Below we respond point-by-point and indicate the revisions we will make.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that ALEE 'exposes persistent gaps in cross-lingual semantic representation' rests on the unverified assumption that translations of the AMR minimal pairs preserve the controlled semantic shifts without introducing artifacts or losing distinctions. No translation-quality controls, back-translation checks, or human fidelity ratings are described, which directly threatens the targeted-diagnostic interpretation, especially for low-resource languages where MT quality is lower.

    Authors: We acknowledge that the current manuscript does not include dedicated translation-quality controls or fidelity ratings for the minimal-pair contrasts. The translations are taken directly from three established parallel corpora rather than generated via MT for the study; however, we did not verify that the fine-grained semantic distinctions survive translation. We will add a new subsection on translation fidelity that reports (1) back-translation consistency on a stratified sample of pairs, (2) automatic semantic similarity scores between English and translated pairs, and (3) a discussion of known limitations for languages with lower-quality parallel data. This will be accompanied by a clearer statement of the assumption and its scope. revision: yes

  2. Referee: [Abstract] Abstract: the reported empirical variation across languages, lengths, and phenomena is presented without any mention of statistical significance testing, data-exclusion criteria, or sensitivity to implementation choices in the translation pipeline, leaving the load-bearing cross-lingual claims difficult to evaluate for robustness.

    Authors: We agree that the absence of statistical testing and explicit robustness checks weakens the presentation. In the revision we will (a) report statistical significance (paired t-tests or Wilcoxon tests with multiple-comparison correction) for the main cross-lingual, length, and phenomenon contrasts, (b) state data-exclusion criteria (minimum number of valid pairs per language, removal of languages with <50 pairs), and (c) add a sensitivity analysis that re-runs key results after excluding the lowest-quality parallel dataset. Because the translation step is fixed by the choice of parallel corpora, the sensitivity analysis will focus on dataset choice rather than an arbitrary MT pipeline. revision: yes

Circularity Check

0 steps flagged

No significant circularity: ALEE is an independent evaluation framework

full rationale

The paper introduces ALEE as a new evaluation framework that extends Sentence Smith (Li et al., 2025) by generating AMR-based English minimal pairs and pairing them with translations. No equations, fitted parameters, or predictions that reduce to inputs by construction are described. The central claims rest on empirical results across 275+ languages rather than any self-definitional, fitted-input, or self-citation load-bearing step. The cited prior work is independent and is being extended, not invoked as a uniqueness theorem or ansatz that forces the result. This is a standard case of an evaluation tool whose validity depends on external assumptions (e.g., translation fidelity) but whose derivation chain does not collapse into its own inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The approach rests on the assumption that AMR can produce controllable semantic minimal pairs and that parallel translations preserve those distinctions; no free parameters or invented entities are mentioned.

axioms (2)
  • domain assumption AMR representations can be used to generate English minimal pairs with fine-grained, controlled semantic shifts
    Central to the method described in the abstract for creating the evaluation pairs.
  • domain assumption Translations into target languages preserve the semantic distinctions introduced in the English AMR pairs
    Required for the cross-lingual diagnostic to be valid; not discussed in abstract.

pith-pipeline@v0.9.1-grok · 5737 in / 1400 out tokens · 23753 ms · 2026-07-02T19:08:49.602657+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

126 extracted references · 71 canonical work pages · 7 internal anchors

  1. [1]

    How Good is Your Tokenizer? On the Monolingual Performance of Multilingual Language Models , author=. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers) , month=. 2021 , address=. doi:10.18653/v1/2021.acl-long.243 , pages=

    work page doi:10.18653/v1/2021.acl-long.243 2021

  2. [2]

    language , volume=

    Thematic proto-roles and argument selection , author=. language , volume=. 1991 , url=

    1991

  3. [3]

    Similar, but why? A Toolkit for Explaining Text Similarity

    Opitz, Juri and Michail, Andrianos and Moeller, Lucas and Pad \'o , Sebastian and Clematide, Simon. Similar, but why? A Toolkit for Explaining Text Similarity. Proceedings of the 19th Conference of the E uropean Chapter of the A ssociation for C omputational L inguistics (Volume 3: System Demonstrations). 2026. doi:10.18653/v1/2026.eacl-demo.16

    work page doi:10.18653/v1/2026.eacl-demo.16 2026

  4. [4]

    Transactions of the Association for Computational Linguistics , volume=

    Jumelet, Jaap and Weissweiler, Leonie and Nivre, Joakim and Bisazza, Arianna , title=. Transactions of the Association for Computational Linguistics , volume=. 2026 , month=. doi:10.1162/TACL.a.600 , url=

    work page doi:10.1162/tacl.a.600 2026

  5. [5]

    A Survey of the State of Explainable

    Danilevsky, Marina and Qian, Kun and Aharonov, Ranit and Katsis, Yannis and Kawas, Ban and Sen, Prithviraj , editor=. A Survey of the State of Explainable. Proceedings of the 1st Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 10th International Joint Conference on Natural Language Processing , month=. 2020 ...

    work page doi:10.18653/v1/2020.aacl-main.46 2020

  6. [6]

    Sentence Analogies:

    Zhu, Xunjie and de Melo, Gerard , editor=. Sentence Analogies:. Proceedings of the 28th International Conference on Computational Linguistics , month=. 2020 , address=. doi:10.18653/v1/2020.coling-main.300 , pages=

    work page doi:10.18653/v1/2020.coling-main.300 2020

  7. [7]

    Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers) , month=

    Exploring Semantic Properties of Sentence Embeddings , author=. Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers) , month=. 2018 , address=. doi:10.18653/v1/P18-2100 , pages=

    work page doi:10.18653/v1/p18-2100 2018

  8. [8]

    2018 , language=

    Manual of Romance Sociolinguistics , author=. 2018 , language=

    2018

  9. [9]

    2004 , language=

    Romansh: Facts & figures , author=. 2004 , language=

    2004

  10. [10]

    Richtlinien für die Gestaltung einer gesamtbündnerromanischen Schriftsprache :

    Schmid, Heinrich , address=. Richtlinien für die Gestaltung einer gesamtbündnerromanischen Schriftsprache :. Richtlinien für die Gestaltung einer gesamtbündnerromanischen Schriftsprache Rumantsch grischun , edition=. 1982 , isbn=

    1982

  11. [11]

    2010 , language=

    Atlas of the world's languages in danger , author=. 2010 , language=

    2010

  12. [12]

    1975 , issue_date=

    Wilks, Yorick , title=. 1975 , issue_date=. doi:10.1145/360762.360770 , journal=

    work page doi:10.1145/360762.360770 1975

  13. [13]

    and Ullman, Jeffrey D

    Aho, Alfred V. and Ullman, Jeffrey D. , title=. 1972 , isbn=

    1972

  14. [14]

    Interpretable Text Embeddings and Text Similarity Explanation:

    Opitz, Juri and Moeller, Lucas and Michail, Andrianos and Pad. Interpretable Text Embeddings and Text Similarity Explanation:. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing , month=. 2025 , address=. doi:10.18653/v1/2025.emnlp-main.1135 , pages=

    work page doi:10.18653/v1/2025.emnlp-main.1135 2025

  15. [15]

    Chandra and Dexter C

    Ashok K. Chandra and Dexter C. Kozen and Larry J. Stockmeyer , year=. Alternation , journal=

  16. [16]

    2007 , isbn=

    Andrew, Galen and Gao, Jianfeng , title=. 2007 , isbn=. doi:10.1145/1273496.1273501 , booktitle=

    work page doi:10.1145/1273496.1273501 2007

  17. [17]

    1997 , publisher=

    Dan Gusfield , title=. 1997 , publisher=

    1997

  18. [18]

    Tetreault , title=

    Mohammad Sadegh Rasooli and Joel R. Tetreault , title=. Computing Research Repository , volume=. 2015 , url=

    2015

  19. [19]

    Journal of Machine Learning Research , year=

    Rie Kubota Ando and Tong Zhang , title=. Journal of Machine Learning Research , year=

  20. [20]

    Sentence Smith:

    Li, Hongji and Michail, Andrianos and Gubelmann, Reto and Clematide, Simon and Opitz, Juri , editor=. Sentence Smith:. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing , month=. 2025 , address=. doi:10.18653/v1/2025.emnlp-main.1343 , pages=

    work page doi:10.18653/v1/2025.emnlp-main.1343 2025

  21. [21]

    2511.14758 , archiveprefix=

    Dawn Lawrie and James Mayfield and Eugene Yang and Andrew Yates and Sean MacAvaney and Ronak Pradeep and Scott Miller and Paul McNamee and Luca Soldani , year=. 2511.14758 , archiveprefix=

    work page arXiv

  22. [22]

    Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics , month=

    Lewis, Patrick and Oguz, Barlas and Rinott, Ruty and Riedel, Sebastian and Schwenk, Holger , editor=. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics , month=. 2020 , address=. doi:10.18653/v1/2020.acl-main.653 , pages=

    work page doi:10.18653/v1/2020.acl-main.653 2020

  23. [23]

    2022 , isbn=

    Bonifacio, Luiz and Abonizio, Hugo and Fadaee, Marzieh and Nogueira, Rodrigo , title=. 2022 , isbn=. doi:10.1145/3477495.3531863 , booktitle=

    work page doi:10.1145/3477495.3531863 2022

  24. [24]

    Proceedings of the 11th International Workshop on Semantic Evaluation (

    Cer, Daniel and Diab, Mona and Agirre, Eneko and Lopez-Gazpio, I. Proceedings of the 11th International Workshop on Semantic Evaluation (. 2017 , address=. doi:10.18653/v1/S17-2001 , pages=

    work page doi:10.18653/v1/s17-2001 2017

  25. [25]

    Overview of the Second

    Zweigenbaum, Pierre and Sharoff, Serge and Rapp, Reinhard , editor=. Overview of the Second. Proceedings of the 10th Workshop on Building and Using Comparable Corpora , month=. 2017 , address=. doi:10.18653/v1/W17-2512 , pages=

    work page doi:10.18653/v1/w17-2512 2017

  26. [26]

    Examining Multilingual Embedding Models Cross-Lingually Through

    Michail, Andrianos and Clematide, Simon and Sennrich, Rico , editor=. Examining Multilingual Embedding Models Cross-Lingually Through. Findings of the Association for Computational Linguistics: EMNLP 2025 , month=. 2025 , address=. doi:10.18653/v1/2025.findings-emnlp.115 , pages=

    work page doi:10.18653/v1/2025.findings-emnlp.115 2025

  27. [27]

    , editor=

    Magomere, Jabez and La Malfa, Emanuele and Tonneau, Manuel and Kazemi, Ashkan and Hale, Scott A. , editor=. When Claims Evolve:. Findings of the Association for Computational Linguistics: ACL 2025 , month=. 2025 , address=. doi:10.18653/v1/2025.findings-acl.1150 , pages=

    work page doi:10.18653/v1/2025.findings-acl.1150 2025

  28. [28]

    No Language Left Behind: Scaling Human-Centered Machine Translation

    NLLB Team and Marta R. Costa-jussà and others , year=. No Language Left Behind:. 2207.04672 , archiveprefix=

    work page internal anchor Pith review Pith/arXiv arXiv

  29. [29]

    Findings of the Association for Computational Linguistics: ACL 2025 , month=

    Deutsch, Daniel and Briakou, Eleftheria and Caswell, Isaac Rayburn and Finkelstein, Mara and Galor, Rebecca and Juraska, Juraj and Kovacs, Geza and Lui, Alison and Rei, Ricardo and Riesa, Jason and Rijhwani, Shruti and Riley, Parker and Salesky, Elizabeth and Trabelsi, Firas and Winkler, Stephanie and Zhang, Biao and Freitag, Markus , editor=. Findings of...

    work page doi:10.18653/v1/2025.findings-acl.634 2025

  30. [30]

    Expanding the

    Vamvas, Jannis and P. Expanding the. Proceedings of the Tenth Conference on Machine Translation , month=. 2025 , address=. doi:10.18653/v1/2025.wmt-1.79 , pages=

    work page doi:10.18653/v1/2025.wmt-1.79 2025

  31. [31]

    International Conference on Learning Representations , editor=

    Enevoldsen, Kenneth and Chung, Isaac and Kerboua, Imene and Kardos, M\'. International Conference on Learning Representations , editor=

  32. [32]

    Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP) , month=

    Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillation , author=. Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP) , month=. 2020 , address=. doi:10.18653/v1/2020.emnlp-main.365 , pages=

    work page doi:10.18653/v1/2020.emnlp-main.365 2020

  33. [33]

    jina-embeddings-v5-text: Task-Targeted Embedding Distillation

    Mohammad Kalim Akram and Saba Sturua and Nastia Havriushenko and Quentin Herreros and Michael Günther and Maximilian Werk and Han Xiao , year=. jina-embeddings-v5-text:. 2602.15547 , archiveprefix=

    work page internal anchor Pith review Pith/arXiv arXiv

  34. [34]

    Multilingual E5 Text Embeddings: A Technical Report

    Liang Wang and Nan Yang and Xiaolong Huang and Linjun Yang and Rangan Majumder and Furu Wei , year=. Multilingual E5 Text Embeddings:. 2402.05672 , archiveprefix=

    work page internal anchor Pith review Pith/arXiv arXiv

  35. [35]

    EmbeddingGemma: Powerful and Lightweight Text Representations

    Henrique Schechter Vera and Sahil Dua and Biao Zhang and Daniel Salz and Ryan Mullins and Sindhu Raghuram Panyam and Sara Smoot and Iftekhar Naim and Joe Zou and Feiyang Chen and Daniel Cer and Alice Lisak and Min Choi and Lucas Gonzalez and Omar Sanseviero and Glenn Cameron and Ian Ballantyne and Kat Black and Kaifeng Chen and Weiyi Wang and Zhe Li and G...

    work page internal anchor Pith review Pith/arXiv arXiv

  36. [36]

    Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: Industry Track , month=

    Zhang, Xin and Zhang, Yanzhao and Long, Dingkun and Xie, Wen and Dai, Ziqi and Tang, Jialong and Lin, Huan and Yang, Baosong and Xie, Pengjun and Huang, Fei and Zhang, Meishan and Li, Wenjie and Zhang, Min , editor=. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: Industry Track , month=. 2024 , address=. doi:10.186...

    work page doi:10.18653/v1/2024.emnlp-industry.103 2024

  37. [37]

    2025 , eprint=

    Training Sparse Mixture Of Experts Text Embedding Models , author=. 2025 , eprint=

    2025

  38. [38]

    2025 , eprint=

    Granite Embedding Models , author=. 2025 , eprint=

    2025

  39. [39]

    Information Representation Fairness in Long-Document Embeddings: The Peculiar Interaction of Positional and Language Bias

    Schuhmacher, Elias and Michail, Andrianos and Opitz, Juri and Sennrich, Rico and Clematide, Simon. Information Representation Fairness in Long-Document Embeddings: The Peculiar Interaction of Positional and Language Bias. Findings of the A ssociation for C omputational L inguistics: ACL 2026. 2026

    2026

  40. [40]

    Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics , month=

    Unsupervised Cross-lingual Representation Learning at Scale , author=. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics , month=. 2020 , address=. doi:10.18653/v1/2020.acl-main.747 , pages=

    work page doi:10.18653/v1/2020.acl-main.747 2020

  41. [41]

    Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing , month=

    Andrews, Pierre and Artetxe, Mikel and Meglioli, Mariano Coria and Costa-juss. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing , month=. 2025 , address=. doi:10.18653/v1/2025.emnlp-main.1400 , pages=

    work page doi:10.18653/v1/2025.emnlp-main.1400 2025

  42. [42]

    Omnilingual

    The Omnilingual MT Team and Belen Alastruey and Niyati Bafna and Andrea Caciolai and Kevin Heffernan and Artyom Kozhevnikov and Christophe Ropers and Eduardo S. Omnilingual. 2026 , eprint=

    2026

  43. [43]

    2025 , eprint=

    Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Models , author=. 2025 , eprint=

    2025

  44. [44]

    M 3-Embedding: Multi-Linguality, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation

    Chen, Jianlyu and Xiao, Shitao and Zhang, Peitian and Luo, Kun and Lian, Defu and Liu, Zheng. M 3-Embedding: Multi-Linguality, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation. Findings of the Association for Computational Linguistics: ACL 2024. 2024. doi:10.18653/v1/2024.findings-acl.137

    work page doi:10.18653/v1/2024.findings-acl.137 2024

  45. [45]

    IEEE Trans

    Fangzhi Xu and Qika Lin and Jiawei Han and Tianzhe Zhao and Jun Liu and Erik Cambria , title=. IEEE Trans. Knowl. Data Eng. , volume=. 2025 , month=

    2025

  46. [46]

    Compositionality and Sentence Meaning: Comparing Semantic Parsing and Transformers on a Challenging Sentence Similarity Dataset

    Fodor, James and De Deyne, Simon and Suzuki, Shinsuke. Compositionality and Sentence Meaning: Comparing Semantic Parsing and Transformers on a Challenging Sentence Similarity Dataset. Computational Linguistics. 2025. doi:10.1162/coli_a_00536

    work page doi:10.1162/coli_a_00536 2025

  47. [47]

    Minji Kim and Whanhee Cho and Soohyeong Kim and Yong Suk Choi , title=. Adv. Intell. Syst. , volume=. 2024 , month=

    2024

  48. [48]

    When Truth Matters - Addressing Pragmatic Categories in Natural Language Inference (

    Gubelmann, Reto and Kalouli, Aikaterini-lida and Niklaus, Christina and Handschuh, Siegfried , editor=. When Truth Matters - Addressing Pragmatic Categories in Natural Language Inference (. Proceedings of the 12th Joint Conference on Lexical and Computational Semantics (*SEM 2023) , month=. 2023 , address=. doi:10.18653/v1/2023.starsem-1.4 , pages=

    work page doi:10.18653/v1/2023.starsem-1.4 2023

  49. [49]

    Interpretable Text Embeddings and Text Similarity Explanation:

    Opitz, Juri and M. Interpretable Text Embeddings and Text Similarity Explanation:. EMNLP 2025 , year=

    2025

  50. [50]

    Proceedings of the 15th International Conference on Computational Semantics , month=

    Opitz, Juri and Wein, Shira and Steen, Julius and Frank, Anette and Schneider, Nathan , editor=. Proceedings of the 15th International Conference on Computational Semantics , month=. 2023 , address=

    2023

  51. [51]

    2013 , address=

    Bhagat, Rahul and Hovy, Eduard , journal=. 2013 , address=. doi:10.1162/COLI_a_00166 , pages=

    work page doi:10.1162/coli_a_00166 2013

  52. [52]

    Paraphrase Generation:

    Zhou, Jianing and Bhat, Suma , editor=. Paraphrase Generation:. Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing , month=. 2021 , address=. doi:10.18653/v1/2021.emnlp-main.414 , pages=

    work page doi:10.18653/v1/2021.emnlp-main.414 2021

  53. [53]

    ArXiv , year=

    Mistral 7B , author=. ArXiv , year=

  54. [54]

    Qwen Technical Report

    Qwen technical report , author=. arXiv preprint arXiv:2309.16609 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  55. [55]

    Towards General Text Embeddings with Multi-stage Contrastive Learning

    Towards general text embeddings with multi-stage contrastive learning , author=. arXiv preprint arXiv:2308.03281 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  56. [56]

    Paraphrasing evades detectors of

    Krishna, Kalpesh and Song, Yixiao and Karpinska, Marzena and Wieting, John and Iyyer, Mohit , booktitle=. Paraphrasing evades detectors of

  57. [57]

    Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing , month=

    A large annotated corpus for learning natural language inference , author=. Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing , month=. 2015 , address=. doi:10.18653/v1/D15-1075 , pages=

    work page doi:10.18653/v1/d15-1075 2015

  58. [58]

    Benchmark Probing:

    Chunyuan Deng and Yilun Zhao and Xiangru Tang and Mark Gerstein and Arman Cohan , booktitle=. Benchmark Probing:. 2024 , url=

    2024

  59. [59]

    Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics , month=

    Syntactic Data Augmentation Increases Robustness to Inference Heuristics , author=. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics , month=. 2020 , address=. doi:10.18653/v1/2020.acl-main.212 , pages=

    work page doi:10.18653/v1/2020.acl-main.212 2020

  60. [60]

    Linguistically-Informed Transformations (

    Li, Chuanrong and Shengshuo, Lin and Liu, Zeyu and Wu, Xinyi and Zhou, Xuhui and Steinert-Threlkeld, Shane , editor=. Linguistically-Informed Transformations (. Proceedings of the Third BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP , month=. 2020 , address=. doi:10.18653/v1/2020.blackboxnlp-1.12 , pages=

    work page doi:10.18653/v1/2020.blackboxnlp-1.12 2020

  61. [61]

    Proceedings of the 62nd

    Parmar, Mihir and Patel, Nisarg and Varshney, Neeraj and Nakamura, Mutsumi and Luo, Man and Mashetty, Santosh and Mitra, Arindam and Baral, Chitta , year=. Proceedings of the 62nd. doi:10.18653/v1/2024.acl-long.739 , urldate=

    work page doi:10.18653/v1/2024.acl-long.739 2024

  62. [62]

    Translate, then Parse! A Strong Baseline for Cross-Lingual

    Uhrig, Sarah and Garcia, Yoalli and Opitz, Juri and Frank, Anette , editor=. Translate, then Parse! A Strong Baseline for Cross-Lingual. Proceedings of the 17th International Conference on Parsing Technologies and the IWPT 2021 Shared Task on Parsing into Enhanced Universal Dependencies (IWPT 2021) , month=. 2021 , address=. doi:10.18653/v1/2021.iwpt-1.6 , pages=

    work page doi:10.18653/v1/2021.iwpt-1.6 2021

  63. [63]

    Palmer, Martha and Gildea, Daniel and Kingsbury, Paul , journal=. The. 2005 , url=. doi:10.1162/0891201053630264 , pages=

    work page doi:10.1162/0891201053630264 2005

  64. [64]

    Proceedings of the 9th Workshop on Representation Learning for NLP (RepL4NLP-2024) , month=

    Tracking linguistic information in transformer-based sentence embeddings through targeted sparsification , author=. Proceedings of the 9th Workshop on Representation Learning for NLP (RepL4NLP-2024) , month=. 2024 , address=

    2024

  65. [65]

    Proceedings of the 10th Italian Conference on Computational Linguistics (CLiC-it 2024) , month=

    Exploring Syntactic Information in Sentence Embeddings through Multilingual Subject-verb Agreement , author=. Proceedings of the 10th Italian Conference on Computational Linguistics (CLiC-it 2024) , month=. 2024 , address=

    2024

  66. [66]

    Exploring

    Nastase, Vivi and Samo, Giuseppe and Jiang, Chunyang and Merlo, Paola , editor=. Exploring. Proceedings of the 10th Italian Conference on Computational Linguistics (CLiC-it 2024) , month=. 2024 , address=

    2024

  67. [67]

    Annotation of Tense and Aspect Semantics for Sentential

    Donatelli, Lucia and Regan, Michael and Croft, William and Schneider, Nathan , editor=. Annotation of Tense and Aspect Semantics for Sentential. Proceedings of the Joint Workshop on Linguistic Annotation, Multiword Expressions and Constructions (. 2018 , address=

    2018

  68. [68]

    Journal of Child Language , volume=

    The development of sentence planning , author=. Journal of Child Language , volume=. 1990 , doi=

    1990

  69. [69]

    , booktitle=

    Kasper, Robert T. , booktitle=. A Flexible Interface for Linking Applications to. 1989 , url=

    1989

  70. [70]

    First Conference on Language Modeling , year=

    Description-Based Text Similarity , author=. First Conference on Language Modeling , year=

  71. [71]

    Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing , month=

    Natural Language Decompositions of Implicit Content Enable Better Text Representations , author=. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing , month=. 2023 , address=. doi:10.18653/v1/2023.emnlp-main.815 , pages=

    work page doi:10.18653/v1/2023.emnlp-main.815 2023

  72. [72]

    Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , month=

    Fan, Lizhou and Hua, Wenyue and Li, Lingyao and Ling, Haoyang and Zhang, Yongfeng , editor=. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , month=. 2024 , address=. doi:10.18653/v1/2024.acl-long.225 , pages=

    work page doi:10.18653/v1/2024.acl-long.225 2024

  73. [73]

    Interfacing an English text generator with a German

    Bateman, John and Kasper, Robert and Sch. Interfacing an English text generator with a German. Interaktion und Kommunikation mit dem Computer: Jahrestagung der Gesellschaft f. 1990 , organization=

    1990

  74. [74]

    Mann and Christian M

    William C. Mann and Christian M. I. M. Matthiessen , title =. 1983 , url =

    1983

  75. [75]

    Strategic Computing - Natural Language Workshop: Proceedings of a Workshop Held at Marina del Rey, California, May 1-2, 1986 , year=

    A Logical-Form and Knowledge-Base Design for Natural Language Generation , author=. Strategic Computing - Natural Language Workshop: Proceedings of a Workshop Held at Marina del Rey, California, May 1-2, 1986 , year=

    1986

  76. [76]

    Opitz, Juri and Frank, Anette , editor=. Better. Proceedings of the 3rd Workshop on Evaluation and Comparison of NLP Systems , month=. 2022 , address=. doi:10.18653/v1/2022.eval4nlp-1.4 , pages=

    work page doi:10.18653/v1/2022.eval4nlp-1.4 2022

  77. [77]

    A Human Evaluation of

    Manning, Emma and Wein, Shira and Schneider, Nathan , editor=. A Human Evaluation of. Proceedings of the 28th International Conference on Computational Linguistics , month=. 2020 , address=. doi:10.18653/v1/2020.coling-main.420 , pages=

    work page doi:10.18653/v1/2020.coling-main.420 2020

  78. [78]

    Modeling Quantification and Scope in

    Pustejovsky, James and Lai, Ken and Xue, Nianwen , editor=. Modeling Quantification and Scope in. Proceedings of the First International Workshop on Designing Meaning Representations , month=. 2019 , address=. doi:10.18653/v1/W19-3303 , pages=

    work page doi:10.18653/v1/w19-3303 2019

  79. [79]

    Ambiguity and Disagreement in

    Wein, Shira , editor=. Ambiguity and Disagreement in. Proceedings of Context and Meaning: Navigating Disagreements in NLP Annotation , month=. 2025 , address=

    2025

  80. [80]

    Proceedings of the 31st International Conference on Computational Linguistics , month=

    Michail, Andrianos and Clematide, Simon and Opitz, Juri , editor=. Proceedings of the 31st International Conference on Computational Linguistics , month=. 2025 , address=

    2025

Showing first 80 references.