Pith. sign in

REVIEW 6 major objections 9 minor 34 references

No single text embedding model wins everywhere: pick by task, latency, and cost, with mE5-L as the practical open-source default and T3EM only when peak retrieval quality justifies API cost.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-30 20:33 UTC pith:JNP7NH3O

load-bearing objection Small honest bake-off plus a usable selection checklist; the science is thin, the engineering advice is mostly sound if you stay inside the four-dataset fence. the 6 major comments →

arxiv 2607.23507 v1 pith:JNP7NH3O submitted 2026-07-26 cs.IR cs.AI

Choosing a Text Embedding Model: A Practical Benchmarking and Decision Framework

Madhav S Baidya This is my paper
classification cs.IR cs.AI
keywords text embeddingsdense retrievalRAGMTEBBEIRdocument chunkingvector databasesANN search
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This report argues that the embedding model topping a leaderboard is rarely the right choice for a real retrieval or search system. On four English retrieval benchmarks, a commercial API model (T3EM) posts the highest average nDCG@10 (0.638), but open-source mE5-L comes within about 0.09 points at roughly 7–14× lower latency and no per-query fee, and is offered as the default when requirements are unspecified. Models trained for sentence similarity rather than retrieval (LaBSE, mMPNet) lag badly on retrieval despite solid similarity scores, showing that training objective outweighs model size. Across the wider MTEB landscape, different families lead different tasks—ST5 on semantic similarity, MPNet on clustering and reranking, GTR/SGPT on broader retrieval, LaBSE on bitext mining. The author also shows that chunk quality plateaus by about 32 tokens and collapses below roughly 16, and that model choice only makes sense inside the full pipeline of chunking, indexing, ANN search, and optional reranking.

Core claim

Embedding model selection is a multi-objective decision, not a search for one leaderboard winner. On the primary English retrieval subsets, T3EM leads (average nDCG@10 = 0.638) but at high latency and API cost; mE5-L is the strongest measured open-source alternative (0.546) at open-source speed and cost, and is the recommended default when requirements are unknown. Training objective—retrieval versus sentence similarity—dominates size for retrieval quality, and no model family wins across all MTEB task types.

What carries the argument

A practical selection framework that ranks models by primary downstream task, acceptable latency, document length/context limits, and query–document wording mismatch, grounded in directly measured nDCG@10 on four BEIR subsets plus cited MTEB task scores and a chunk-size ablation (quality plateaus by 32 tokens; collapses below ~16).

Load-bearing premise

That rankings and the mE5-L default recommendation generalize from only four English retrieval datasets plus unre-measured published MTEB scores, and that isolated retrieval scores are enough to guide real deployments even though end-to-end answer quality was not measured.

What would settle it

Re-run the same six models on a broader held-out English (and non-English) retrieval suite under one pipeline; if mE5-L no longer sits near T3EM on nDCG@10, or if a similarity-trained model matches retrieval-trained models, the default recommendation and the training-objective claim fail. Separately, measure end-to-end RAG answer quality: if higher nDCG@10 does not improve final answers under fixed reranker and generator, retrieval-only ranking is not a sufficient proxy.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • For unspecified English RAG requirements, start with mE5-L rather than the commercial API or a similarity-trained model.
  • Move to T3EM (or another long-context retrieval model) only when peak quality, long documents, or large query–document vocabulary mismatch justify latency and API cost.
  • Do not repurpose paraphrase/similarity models (e.g. LaBSE, mMPNet) as primary retrievers; choose STS-, clustering-, or bitext-specialized families for those tasks instead.
  • Keep chunks at least ~16 tokens and treat ~32 tokens as the practical plateau; prefer semantic or recursive splitting mainly at small sizes.
  • Treat embedding choice as one stage in chunking → encode → ANN index → optional rerank, not as an isolated leaderboard pick.

Where Pith is reading between the lines

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

  • Teams that re-index large corpora often will feel API cost and p95 latency more than per-query benchmarks suggest, pushing the default even harder toward self-hosted open models unless quality gaps are large on their own data.
  • If newer mid-size open models (estimated near T3EM on BEIR averages) hold up under the same primary subsets, the commercial quality edge may shrink to long-context and asymmetric-training niches rather than raw English retrieval.
  • Parent-child and hierarchical chunking, left untested here, are the natural next ablation for long structured documents where 32-token plateaus may not capture the full context the generator needs.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

6 major / 9 minor

Summary. The manuscript is a practitioner-oriented benchmarking report comparing a commercial API embedding model (pseudonymized as "T3EM") against five open-source models on four English BEIR retrieval subsets (FiQA-2018, NFCorpus, SciFact, TREC-COVID), supplemented by (a) latency and cost measurements, (b) a small chunking ablation, (c) a cited-not-remeasured synthesis of MTEB results across seven other task categories, and (d) a decision framework with task- and constraint-conditioned model recommendations. Central claims: T3EM achieves the highest average nDCG@10 (0.638) among directly measured models; mE5-L is the strongest open-source alternative (0.546) at ~7–14× lower latency and is recommended as the default; training objective (retrieval vs. sentence-similarity) dominates model size for retrieval quality; chunking quality plateaus by 32 tokens and collapses below ~16; and no single model leads across all MTEB task types.

Significance. If the reported numbers hold, the paper's useful contributions are: a directly measured, same-pipeline comparison of a commercial API embedder against common open-source baselines, including latency and per-token cost figures that are rarely reported jointly; an honest and specific limitations section (§7.8) that discloses the four-dataset scope, cited MTEB scores, and environment-specific latency; and a consolidated, task-conditioned selection framework that practitioners will find readable. The narrow claim (Table 19 ordering on four BEIR subsets) is falsifiable and plausible. However, the significance is bounded by the small, English-only evaluation scope, the absence of uncertainty quantification, missing evaluation-protocol detail, and an MTEB "landscape" drawn from the 2022-era model pool. The chunking ablation could be a genuine contribution but is currently reported without data.

major comments (6)
  1. [§6.1, Table 19; §7.4, §7.7] Two problems in the central results table. (1) The BGE-M3 row has no TREC-COVID entry ('—'), yet its 'Average' (0.437) is computed over three datasets while every other row averages four; the column is therefore not comparable across rows, and the §7.6 claim that BGE-M3 'underperformed the simpler E5-large' rests on this. Complete the cell, or annotate the column and report like-for-like averages. (2) No uncertainty quantification is reported anywhere. The mE5-L vs. E5-large gap (0.546 vs. 0.538) is 0.008 nDCG over 1,521 queries (TREC-COVID has only 50), yet §7.7's default recommendation and §7.4's 'narrowly ahead' hinge on it, and §7.2 asserts 'statistically comparable' quality for some open-source models with no test performed. Report per-dataset confidence intervals or a paired bootstrap over queries before making pairwise recommendations on sub-0.01 margins.
  2. [§6.2, Table 20; Executive Summary] The '7–14× latency' claim and the latency-conditioned recommendations rest on measurements with unspecified hardware, batch size, warmup, number of runs, and network conditions. The near-identical medians (30.9/30.9/31.0 ms; 16.6/16.6 ms) suggest coarse timing or very few samples. Whether local models were CPU- or GPU-served is not stated, even though §7.5 itself warns that CPU latency does not transfer to GPU deployments — which undercuts the deployment relevance of Table 20 as printed. Specify the environment, run counts, and variance, and clarify what the API latency includes (network round trip, queueing).
  3. [§6.3] The chunking conclusions (all models reach 95% of peak nDCG@10 by 32 tokens; semantic chunking beats fixed-size by 0.090/0.075 at 16 tokens; collapse below 16) are stated in prose with no table, figure, or methodology. BEIR documents (e.g., SciFact abstracts) are far longer than 32 tokens, so the procedure — how documents were chunked, how chunk-level retrieval scores were aggregated to document-level nDCG@10 against document-level qrels, and what 'peak' and '95% of peak' denote — must be specified and the underlying numbers shown. As printed, these claims are unverifiable and run counter to common practice (128–512-token chunks), so they need evidence, not just disclosure in §7.8 that other strategies were untested.
  4. [§6.5–6.6, Table 22] The 'Broader MTEB Performance Summary' and 'Best Performing Models' present the original 2022-era MTEB-paper model pool as the current landscape, omitting the E5/BGE/GTE/Qwen3 families — including models the paper itself lists in §4.3 with estimated scores (Qwen3-Embedding-4B at ∼0.63–0.66) that would top Table 22. The MTEB version/snapshot is never stated. Either label these tables explicitly as the original MTEB-paper pool (not the current leaderboard) or update the pool; as framed, 'Best Performing Models' (rank 1: ST5-XXL, 59.51) is misleading for a 2026 readership and sits uneasily next to §4.3.
  5. [Executive Summary; §7.1 vs. §6.7.3] The Executive Summary states 'ST5 leads on semantic similarity' and §7.1 explains 'Why ST5 dominates semantic textual similarity,' but the paper's own §6.7.3 shows MPNet or MPNet-Multilingual as the best model on 9 of 10 STS datasets, with ST5 winning none individually (SGPT-5.8B-NLI takes STS13). If the intended claim is about average STS score across datasets, the averages must be shown; otherwise the text contradicts the paper's own tables and should be corrected.
  6. [§4.1, §6.1; §7.3.3, §7.4] The evaluation protocol behind Table 19 is under-specified: E5-family models require 'query:'/'passage:' prefixes, and scores are sensitive to prefix handling, normalization, exact vs. ANN search, and (for T3EM) the API version, access date, and embedding dimension used. None of these are reported, so Table 19 cannot be reproduced or sanity-checked; adding one reference row comparing an in-house score to the model's published MTEB/BEIR score would validate the pipeline. Relatedly, Nomic-Embed-Text-v1.5 is recommended for long-document retrieval (§7.3.3, §7.4) despite never being measured (§4.3 is estimates only) and despite no long-document dataset appearing in the evaluation — that recommendation is unsupported by the study's own data and should be dropped, hedged, or measured.
minor comments (9)
  1. [Throughout] Table numbering is broken: Tables 1, 3–16, 18, and 21 never appear; the §6.5 table is unnumbered while §6.6 is Table 22. Renumber consecutively.
  2. [Throughout] LaTeX spacing artifacts appear in many places ('F AISS', 'T ransformer', 'V ector search', 'T op-k', 'F ull F orm'), and several URLs in §3 and §4 are broken by spaces ('https : / / huggingface . co / ...'). Please clean up.
  3. [Glossary] 'V-Measure — Validity Measure' is incorrect: V-measure (Rosenberg & Hirschberg, 2007) is not an acronym for 'validity measure.' Correct the expansion and add the citation.
  4. [§4.1 vs. §4.2.6] The anonymization policy is inconsistent: 'T3EM' is pseudonymized while 'OpenAI Ada Similarity' is named. If T3EM is a real commercial product, name the vendor, model version, and access date so the headline comparison is verifiable; if pseudonymization is required, say why.
  5. [§6.7] Clarify that the per-dataset 'best model' pool in §6.7 differs from the Table 19 pool, to avoid reader confusion (e.g., FiQA2018 'best' is MPNet at 49.96 in §6.7.1 while Table 19 reports T3EM at 0.582 on the same dataset).
  6. [§1.4, RQ(1)] Research question (1) asks whether a longer context window and asymmetric encoding provide a measurable advantage, but the design never isolates context length (all four evaluation corpora are short-passage) and never ablates asymmetric encoding. Either add a long-document evaluation or reframe the RQ to match what the study actually tests.
  7. [§2.3] The vector-database recommendations (Qdrant as default for small/mid scale, Milvus at large scale) are asserted without citation or benchmark support; either cite evidence or soften to opinion.
  8. [§4.3] The estimated nDCG ranges lack per-model sources; cite the specific leaderboard snapshot and access date for each estimate, and note that 'BEIR-style average' estimates are not directly comparable to the four-subset average in Table 19.
  9. [Figure 1 caption] The caption references 'Section 2.5' for chunking, but chunking is §2.4; there is no §2.5.

Circularity Check

0 steps flagged

No circularity: empirical benchmark report and practitioner synthesis, not a derivation that redefines its target from fitted inputs or self-citation.

full rationale

The paper’s load-bearing claims are directly measured nDCG@10, latency, and chunking results on four BEIR subsets (Table 19, §6.1–6.3) plus openly cited external MTEB leaderboard standings (§6.5–6.7), with estimated scores for unrun models explicitly labeled as estimates (§4.3, §6.4) and limitations stated (§7.8). Recommendations (mE5-L default; T3EM when quality/context dominate) follow from those reported numbers and cost/latency tables, not from parameters fitted to the same quantities being ‘predicted,’ nor from self-defined uniqueness theorems. Author list is a single name with no load-bearing self-citations in the reference chain; MTEB/BEIR and model papers are third-party. Post-hoc training-objective narratives in §7.1 explain observed specialization but do not close a definitional loop. Dependence on external published tables is ordinary evidence use, not Eq.X≡Eq.Y circularity. Score 0 is appropriate.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 1 invented entities

This is an empirical systems/benchmarking report. It does not introduce physical entities or fitted laws; load-bearing commitments are evaluation conventions (nDCG@10 as primary quality), the representativeness of four English BEIR subsets, treat-as-given published MTEB scores, and black-box access to T3EM. Free parameters are experimental knobs (chunk sizes, latency percentiles) rather than constants fitted to prove a theory.

free parameters (4)
  • Primary retrieval dataset subset (4 of 15 BEIR/MTEB retrieval sets) = 4 datasets
    FiQA, NFCorpus, SciFact, TREC-COVID alone drive the headline average and default recommendation; choice is author-selected, not derived.
  • Chunk size grid and 95%-of-peak threshold = 16 / 32 / 64 / 128 tokens; 95% peak rule
    Claims that quality plateaus by 32 tokens and collapses below ~16 depend on the tested sizes and the '95% of peak nDCG@10' cutoff in §6.3.
  • Latency summary statistics (median, p95) on unspecified hardware/network = T3EM median 231.6 ms vs ~16–31 ms open-source
    Absolute ms figures and 7–14× comparisons are environment-specific (§6.2, §7.8) and function as reporting choices for the cost-quality narrative.
  • Estimated BEIR-style nDCG ranges for unmeasured models = model-specific ~ranges
    §4.3 ranges (e.g. Qwen3-Embedding-4B ~0.63–0.66) are hand-set from public leaderboards and inform side commentary about matching T3EM.
axioms (5)
  • domain assumption nDCG@10 on BEIR-style retrieval is the primary indicator of embedding quality for ranking/RAG model selection
    Stated in §5.3 and used throughout §6–7; paper notes Recall@k matters for RAG but still centers nDCG@10 for comparability.
  • domain assumption Published MTEB leaderboard scores are comparable enough to own pipeline results to support cross-task winner tables without re-measurement
    §1.1, §6.5–6.7, limitation §7.8 explicitly cite rather than re-verify MTEB.
  • domain assumption Cosine similarity on pooled transformer embeddings plus ANN indexes is the operative retrieval stack being optimized
    Pipeline in §2.1–2.3; sparse/multi-vector mentioned but not the primary measured setting except via BGE-M3 discussion.
  • standard math Standard IR metric definitions (nDCG, MRR, MAP, V-measure, Spearman/Pearson) as formalized in §5.4
    Conventional formulas; not re-derived beyond restatement.
  • ad hoc to paper T3EM API behavior during the study is a stable, representative commercial embedding endpoint
    Model listed without public URL/version (§4.1); entire quality-vs-cost argument depends on that black box.
invented entities (1)
  • T3EM (Text 3 Embedding Model) no independent evidence
    purpose: Commercial API baseline claimed to top the author’s English retrieval averages and anchor when to pay for an API vs open weights.
    Name and results appear specific to this report; no independent model card, paper, or link is given in §4.1, so the entity is not verifiable outside the author’s API access window.

pith-pipeline@v1.2.0-grok45-kimik3 · 26014 in / 4289 out tokens · 91718 ms · 2026-07-30T20:33:53.159638+00:00 · methodology

0 comments
read the original abstract

Choosing the right text embedding model is one of the most consequential -- and most frequently under-examined -- decisions in building a retrieval or search system, yet the model that tops a leaderboard is rarely the best choice for a given deployment. This report develops a practical, evidence-based framework for embedding model selection, built on a benchmarking study that evaluates T3EM (Text 3 Embedding Model), a commercial API-based embedding model, against a broad set of open-source alternatives on English-language retrieval tasks, and situates these findings within the wider Massive Text Embedding Benchmark (MTEB) landscape spanning classification, clustering, semantic similarity, reranking, pair classification, bitext mining, and summarization. Beyond raw benchmark scores, the report traces the full path from embedding model to retrieved result -- how embeddings are produced, how they are indexed and searched at scale, and how document chunking strategy shapes retrieval quality -- so that model choice can be reasoned about as one decision within a complete retrieval pipeline rather than in isolation. The result is a consolidated set of practical recommendations for selecting an embedding model according to task, latency, cost, and deployment constraints.

Figures

Figures reproduced from arXiv: 2607.23507 by Madhav S Baidya.

Figure 1
Figure 1. Figure 1: End-to-end architecture of a text embedding and retrieval system, from raw data sources [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗

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

34 extracted references · 10 linked inside Pith

  1. [1]

    MTEB: Massive Text Embedding Bench- mark,

    N. Muennighoff, N. Tazi, L. Magne, and N. Reimers, “MTEB: Massive Text Embedding Bench- mark,”arXiv preprint arXiv:2210.07316, 2022

    Pith/arXiv arXiv 2022

  2. [2]

    BEIR: A Heterogeneous Bench- mark for Zero-shot Evaluation of Information Retrieval Models,

    N. Thakur, N. Reimers, A. R¨ uckl´ e, A. Srivastava, and I. Gurevych, “BEIR: A Heterogeneous Bench- mark for Zero-shot Evaluation of Information Retrieval Models,”arXiv preprint arXiv:2104.08663, 2021

    Pith/arXiv arXiv 2021

  3. [3]

    WWW’18 Open Challenge: Financial Opinion Mining and Question Answering,

    M. Maia et al., “WWW’18 Open Challenge: Financial Opinion Mining and Question Answering,” inCompanion Proceedings of the The Web Conference 2018, 2018

    2018

  4. [4]

    A Full-Text Learning to Rank Dataset for Medical Information Retrieval,

    V. Boteva, D. Gholipour, A. Sokolov, and S. Riezler, “A Full-Text Learning to Rank Dataset for Medical Information Retrieval,” inAdvances in Information Retrieval (ECIR), 2016

    2016

  5. [5]

    Fact or Fiction: Verifying Scientific Claims,

    D. Wadden, S. Lin, K. Lo, L. L. Wang, M. van Zuylen, A. Cohan, and H. Hajishirzi, “Fact or Fiction: Verifying Scientific Claims,” inProceedings of EMNLP, 2020

    2020

  6. [6]

    TREC-COVID: Constructing a Pandemic Information Retrieval Test Collec- tion,

    E. Voorhees et al., “TREC-COVID: Constructing a Pandemic Information Retrieval Test Collec- tion,”ACM SIGIR Forum, vol. 54, no. 1, 2021

    2021

  7. [7]

    Cumulated Gain-Based Evaluation of IR Techniques,

    K. J¨ arvelin and J. Kek¨ al¨ ainen, “Cumulated Gain-Based Evaluation of IR Techniques,”ACM Trans- actions on Information Systems, vol. 20, no. 4, pp. 422–446, 2002

    2002

  8. [8]

    Text Embeddings by Weakly-Supervised Contrastive Pre-training,

    L. Wang, N. Yang, X. Huang, B. Jiao, L. Yang, D. Jiang, R. Majumder, and F. Wei, “Text Embeddings by Weakly-Supervised Contrastive Pre-training,”arXiv preprint arXiv:2212.03533, 2022

    Pith/arXiv arXiv 2022

  9. [9]

    Multilingual E5 Text Embed- dings: A Technical Report,

    L. Wang, N. Yang, X. Huang, L. Yang, R. Majumder, and F. Wei, “Multilingual E5 Text Embed- dings: A Technical Report,”arXiv preprint arXiv:2402.05672, 2024

    Pith/arXiv arXiv 2024

  10. [10]

    C-Pack: Packaged Resources To Advance General Chinese Embedding,

    S. Xiao, Z. Liu, P. Zhang, and N. Muennighoff, “C-Pack: Packaged Resources To Advance General Chinese Embedding,”arXiv preprint arXiv:2309.07597, 2023

    Pith/arXiv arXiv 2023

  11. [11]

    BGE M3-Embedding: Multi-Lingual, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation,

    J. Chen, S. Xiao, P. Zhang, K. Luo, D. Lian, and Z. Liu, “BGE M3-Embedding: Multi-Lingual, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation,” arXiv preprint arXiv:2402.03216, 2024

    Pith/arXiv arXiv 2024

  12. [12]

    Language-agnostic BERT Sentence Embedding,

    F. Feng, Y. Yang, D. Cer, N. Arivazhagan, and W. Wang, “Language-agnostic BERT Sentence Embedding,” inProceedings of ACL, 2022

    2022

  13. [13]

    MPNet: Masked and Permuted Pre-training for Language Understanding,

    K. Song, X. Tan, T. Qin, J. Lu, and T.-Y. Liu, “MPNet: Masked and Permuted Pre-training for Language Understanding,” inAdvances in Neural Information Processing Systems (NeurIPS), 2020

    2020

  14. [14]

    Sentence-BERT: Sentence Embeddings using Siamese BERT- Networks,

    N. Reimers and I. Gurevych, “Sentence-BERT: Sentence Embeddings using Siamese BERT- Networks,” inProceedings of EMNLP-IJCNLP, 2019

    2019

  15. [15]

    SimCSE: Simple Contrastive Learning of Sentence Embeddings,

    T. Gao, X. Yao, and D. Chen, “SimCSE: Simple Contrastive Learning of Sentence Embeddings,” inProceedings of EMNLP, 2021

    2021

  16. [16]

    Unsuper- vised Dense Information Retrieval with Contrastive Learning,

    G. Izacard, M. Caron, L. Hosseini, S. Riedel, P. Bojanowski, A. Joulin, and E. Grave, “Unsuper- vised Dense Information Retrieval with Contrastive Learning,”Transactions on Machine Learning Research, 2022

    2022

  17. [17]

    SPECTER: Document-level Representation Learning using Citation-informed Transformers,

    A. Cohan, S. Feldman, I. Beltagy, D. Downey, and D. S. Weld, “SPECTER: Document-level Representation Learning using Citation-informed Transformers,” inProceedings of ACL, 2020. Technical ReportChoosing a Text Embedding Model 33

    2020

  18. [18]

    Large Dual Encoders Are Generalizable Retrievers,

    J. Ni et al., “Large Dual Encoders Are Generalizable Retrievers,”arXiv preprint arXiv:2112.07899, 2021

    Pith/arXiv arXiv 2021

  19. [19]

    Sentence-T5: Scalable Sentence Encoders from Pre-trained Text-to-Text Models,

    J. Ni, G. H. ´Abrego, N. Constant, J. Ma, K. B. Hall, D. Cer, and Y. Yang, “Sentence-T5: Scalable Sentence Encoders from Pre-trained Text-to-Text Models,”arXiv preprint arXiv:2108.08877, 2021

    Pith/arXiv arXiv 2021

  20. [20]

    SGPT: GPT Sentence Embeddings for Semantic Search,

    N. Muennighoff, “SGPT: GPT Sentence Embeddings for Semantic Search,”arXiv preprint arXiv:2202.08904, 2022

    Pith/arXiv arXiv 2022

  21. [21]

    GloVe: Global Vectors for Word Representation,

    J. Pennington, R. Socher, and C. D. Manning, “GloVe: Global Vectors for Word Representation,” inProceedings of EMNLP, 2014

    2014

  22. [22]

    Dependency Based Embeddings for Sentence Classification Tasks,

    A. Komninos and S. Manandhar, “Dependency Based Embeddings for Sentence Classification Tasks,” inProceedings of NAACL-HLT, 2016

    2016

  23. [23]

    BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding,

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding,” inProceedings of NAACL-HLT, 2019

    2019

  24. [24]

    Massively Multilingual Sentence Embeddings for Zero-Shot Cross- Lingual Transfer and Beyond,

    M. Artetxe and H. Schwenk, “Massively Multilingual Sentence Embeddings for Zero-Shot Cross- Lingual Transfer and Beyond,”Transactions of the Association for Computational Linguistics, vol. 7, 2019

    2019

  25. [25]

    ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT,

    O. Khattab and M. Zaharia, “ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT,” inProceedings of ACM SIGIR, 2020

    2020

  26. [26]

    SPLADE: Sparse Lexical and Expansion Model for First Stage Ranking,

    T. Formal, B. Piwowarski, and S. Clinchant, “SPLADE: Sparse Lexical and Expansion Model for First Stage Ranking,” inProceedings of ACM SIGIR, 2021

    2021

  27. [27]

    The Probabilistic Relevance Framework: BM25 and Beyond,

    S. Robertson and H. Zaragoza, “The Probabilistic Relevance Framework: BM25 and Beyond,” Foundations and Trends in Information Retrieval, vol. 3, no. 4, pp. 333–389, 2009

    2009

  28. [28]

    Efficient and Robust Approximate Nearest Neighbor Search Using Hierarchical Navigable Small World Graphs,

    Y. A. Malkov and D. A. Yashunin, “Efficient and Robust Approximate Nearest Neighbor Search Using Hierarchical Navigable Small World Graphs,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 42, no. 4, pp. 824–836, 2018

    2018

  29. [29]

    Product Quantization for Nearest Neighbor Search,

    H. J´ egou, M. Douze, and C. Schmid, “Product Quantization for Nearest Neighbor Search,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 33, no. 1, pp. 117–128, 2011

    2011

  30. [30]

    Billion-scale Similarity Search with GPUs,

    J. Johnson, M. Douze, and H. J´ egou, “Billion-scale Similarity Search with GPUs,”IEEE Transac- tions on Big Data, 2019

    2019

  31. [31]

    Milvus: A Purpose-Built Vector Data Management System,

    J. Wang et al., “Milvus: A Purpose-Built Vector Data Management System,” inProceedings of ACM SIGMOD, 2021

    2021

  32. [32]

    Nomic Embed: Training a Reproducible Long Context Text Embedder,

    Z. Nussbaum, J. Morris, B. Duderstadt, and A. Mulyar, “Nomic Embed: Training a Reproducible Long Context Text Embedder,”arXiv preprint arXiv:2402.01613, 2024

    Pith/arXiv arXiv 2024

  33. [33]

    Qdrant: Vector Database for the Next Generation of AI Applications,

    Qdrant, “Qdrant: Vector Database for the Next Generation of AI Applications,”https://qdrant. tech, accessed 2026

    2026

  34. [34]

    Pinecone Vector Database,

    Pinecone Systems, Inc., “Pinecone Vector Database,”https://www.pinecone.io, accessed 2026

    2026