Pith. sign in

REVIEW 4 major objections 5 minor 8 cited by

jina-embeddings-v4: Universal Embeddings for Multimodal Multilingual Retrieval

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a single 3.8-billion-parameter multimodal model, trained once with one joint loss, gives state-of-the-art retrieval for text, code, and visually rich documents in both single-vector and multi-vector output modes.

desk verdict Solid systems contribution with a real benchmark-confounding problem: the headline Jina-VDR SOTA claim is not interpretable without a contamination check, but the model and external-benchmark results are worth engaging. read the letter →

arxiv 2506.18902 v3 pith:2KSABC4P submitted 2025-06-23 cs.AI cs.CLcs.IR

Michael Günther , Saba Sturua , Mohammad Kalim Akram , Isabelle Mohr , Andrei Ungureanu , Bo Wang , Sedigheh Eslami , Scott Martens
show 3 more authors
Maximilian Werk Nan Wang Han Xiao
This is my paper · ORCID
classification cs.AIcs.CLcs.IR
keywords jina-embeddings-v4multimodalembeddingsvisualdocumentretrievallateinteractionmultilingualcodeLoRAadapterscross-modal
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

jina-embeddings-v4 is presented as a single model that can replace the usual stack of a text embedder, an image embedder, and an OCR-plus-retrieval pipeline: one 3.8-billion-parameter vision-language backbone maps text and images into one semantic space, and three small task-specific adapters tailor it to retrieval, semantic similarity, and code search. The core claim is that this one model is the current state of the art for retrieving visually rich documents—screenshots, tables, charts, maps, and mixed-media pages—while staying competitive with specialized text and code embedders. To demonstrate this, the paper introduces Jina-VDR, a multilingual benchmark of more than thirty visual-retrieval tasks, and reports the highest scores on it and on the existing ViDoRe benchmark, with further gains when using multi-vector and late-interaction embeddings. If the claim holds, the practical consequence is that one model, one index, and one embedding space can serve text retrieval, cross-modal search, and visual document search without per-task pipelines.

What carries the argument

The load-bearing object is a unified multimodal language model: images are converted by a vision encoder into a sequence of tokens that the language-model decoder processes exactly like text tokens, so both modalities share contextual attention and the same embedding space. On top are three mechanisms: mean pooling that produces 2048-dimensional single-vector embeddings, trained with a Matryoshka-style loss so they can be truncated to 128 dimensions; a projection layer that produces 128-dimensional per-token multi-vector embeddings, scored by late interaction, that is, summing over query tokens their maximum cosine similarity to any document token; and three 60-million-parameter LoRA adapters for asymmetric retrieval, symmetric text matching, and code search, chosen at inference time. The joint training loss combines InfoNCE contrastive losses on dense and late-interaction similarity matrices plus a KL term that distills late-interaction scores into dense scores, so a single training run teaches both output modes.

What would settle it

Build an independent visual-document retrieval benchmark from documents and human queries created after the model's training data cutoff, with no author involvement, and rerun the reported comparisons; if the margin over the strongest prior visual retrievers narrows or reverses, the state-of-the-art claim is an artifact of benchmark overlap or synthetic-query style.

Watch

Extended reading notes

Core claim

The central discovery is that the modality gap of dual-encoder models is not inevitable: if images are first turned into token sequences by a vision encoder and then passed through the same transformer decoder as text, the model uses a shared embedding region for both modalities. Trained with a joint contrastive loss over dense and late-interaction similarity matrices, the model reaches cross-modal alignment scores around 0.71–0.72 on standard image-text datasets, roughly double a dual-encoder baseline, and reports 84.11 (dense) and 90.17 (late interaction) average nDCG on ViDoRe, and 73.98 and 80.55 on Jina-VDR, the highest scores in the comparison. The paper states that this makes jina-embeddings-v4 the current state of the art, in both single-vector and multi-vector mode, for visually rich document retrieval.

Load-bearing premise

The visual-retrieval state-of-the-art claim stands on the assumption that Jina-VDR is a fair, uncontaminated benchmark, even though the same authors built it and trained the retrieval adapter on some of the same or closely related multimodal data, including VDR multilingual and other curated hard negatives.

Editorial extensions

If this is right

  • Visual document search becomes a single forward pass on the rendered image: no OCR, no layout parsing, and no separate text-extraction pipeline.
  • One frozen backbone with small adapters covers asymmetric retrieval, symmetric similarity, and code search; adding a new task or domain means training a 60-million-parameter adapter, not a 3.8-billion-parameter model.
  • Multi-vector and late-interaction mode is consistently more accurate than single-vector on visual documents (80.55 vs 73.98 average on Jina-VDR), so applications can trade storage and computation for precision by switching output modes.
  • Because text, image, and code embeddings live in the same space, a single index can serve text-to-text, text-to-image, and image-to-text queries without realignment or re-ranking across separate spaces.
  • If Jina-VDR is adopted, visual retrieval evaluation shifts from English and French question answering to multilingual, non-question queries over maps, diagrams, advertisements, and mixed-media pages.

Reading between the lines

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

  • Editorial inference: the benchmark's own LLM-generated and template-based queries may understate the difficulty of real user queries; the natural follow-up is a third-party held-out set of human-typed queries on the same documents, which could be built without retraining the model.
  • Editorial inference: because dense scores are distilled from late-interaction scores, the single-vector mode likely inherits late-interaction's sensitivity to query phrasing; one can test whether truncating dense vectors to 128 dimensions preserves late-interaction rankings better than it preserves dense cosine rankings.
  • Editorial inference: the recipe is not modality-specific; audio, video, or 3D data could in principle be tokenized into the same shared decoder and trained with the same dense-plus-late joint loss, suggesting a path to one universal embedder rather than per-modality models.
  • Editorial inference: if the shared-encoder design is what shrinks the modality gap, then improving the backbone's multilingual text coverage should transfer directly to cross-modal retrieval in low-resource languages; measuring image-to-text retrieval in languages absent from the current evaluation would separate backbone coverage from alignment quality.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper introduces jina-embeddings-v4, a 3.8-billion-parameter multimodal embedding model built on Qwen2.5-VL-3B-Instruct. The model uses task-specific LoRA adapters for asymmetric retrieval, text matching, and code search, and supports both single-vector (Matryoshka-truncatable 2048-dim) and multi-vector late-interaction outputs. Training uses a joint InfoNCE/CoSENT/MRL objective on text-text and text-image pairs. The paper also introduces Jina-VDR, a benchmark extending ViDoRe with 30 additional visually rich document retrieval tasks, and evaluates the model on Jina-VDR, ViDoRe, CLIP Benchmark, MTEB/MMTEB, LongEmbed, CoIR, and STS benchmarks.

Significance. If the results hold, the model provides a practical unified text-image embedding solution, and the Jina-VDR benchmark could become a useful resource for visually rich retrieval. The paper clearly describes the architecture and training procedure, and the released benchmark and model links (Hugging Face) enable reproducibility. The strongest evidence is on ViDoRe and Jina-VDR visual retrieval, where the model outperforms ColPali, DSE, and CLIP-style baselines by large margins. However, the headline 'state-of-the-art' claim is broader than the evidence: on several text benchmarks the model trails gemini-embedding-001 and voyage-code, and the Jina-VDR margin is vulnerable to train/test overlap because the benchmark reuses datasets named in the training section. The paper's own Section 7.1 language ('broadly comparable') is more accurate than the abstract.

major comments (4)
  1. [Abstract; Table 3] The abstract's unqualified claim that jina-embeddings-v4 'achieves state-of-the-art performance on both single-modal and cross-modal retrieval tasks' is contradicted by Table 3. On MMTEB the model scores 66.49 versus gemini-embedding-001's 67.71; on MTEB-en it scores 55.97 versus 64.35; on COIR it scores 71.59 versus voyage-code-3's 77.33 and gemini's 73.11; on LEMB it scores 67.11 versus 78.35. The model is best-in-class only on STS-en and on the visual benchmarks. Please revise the abstract and Section 7.1 to say 'competitive' or 'state-of-the-art on visually rich retrieval' rather than a general SOTA claim.
  2. [Section 5.2.1; Section 6.1; Table A1; Table 3] The central visual-retrieval claim (Section 7.3: 'provides the current state-of-the-art, in both single- and multi-vector mode' on Jina-VDR) is not interpretable without a contamination analysis. Section 5.2.1 states that multimodal hard negatives were mined from 'VDR multilingual' and other curated multimodal datasets, while Section 6.1 lists re-purposed datasets (ChartQA, PlotQA, CharXiv, TableVQA, MPMQA, etc.) that appear as Jina-VDR tasks in Table A1. The paper reports no overlap statistics between training data and Jina-VDR test tasks, no evaluation code, and no variance or significance estimates. If the same documents appear on both sides of the training/test boundary, the margins in Table 3 (e.g., dense 73.98 vs. 67.25 for dse-qwen2-2b-mrl-v1) could partially reflect memorization. Please provide a per-task overlap analysis and report results with and without overlapping tasks, or restrict the SOTA claim to tasks verified to be disjoint.
  3. [Section 6.3] Section 6.3 states that many Jina-VDR queries were generated with Qwen2 and filtered by LLMs, while jina-embeddings-v4 is built on Qwen2.5-VL. Because the query distribution may be systematically closer to the model's prior, the benchmark may overstate the proposed model's advantage over baselines. Please report a sensitivity analysis that excludes LLM-generated/synthetic tasks, or at least report per-task results separately for human-annotated, re-purposed, and synthetic subsets.
  4. [Table A2 and Table A1] Several Jina-VDR tasks are very small (e.g., ramen_benchmark_jp has 29 queries, shanghai_master_plan has 57, jina_2024_yearly_book has 75, student-enrollment has 1000 but near-random scores), so the averaged nDCG@5 numbers in Table A2 and Table 3 are unstable. Reporting no confidence intervals or significance tests makes it difficult to judge whether the 73.98 vs. 67.25 dense margin is real. Please add bootstrap confidence intervals or per-task significance tests, and consider weighting by task size in the average.
minor comments (5)
  1. [Table 3] The model name 'jina-embedings-v2-code' is a typo for 'jina-embeddings-v2-code'; please fix it.
  2. [Section 6] The sentence 'These tests include re-purposed existing datasets, new manually-annotated datasets, and generated synthetic data' is missing a period at the end.
  3. [Section 8.2] The cross-modal alignment score definition should state whether the cosine similarities are averaged over the 1K samples per dataset or over all pairs; Table 4 would benefit from a standard deviation.
  4. [Section 5.1] The notation in Equation (2) says qi,pj∈B, but B is a batch of pairs; please clarify that the division is by the number of query tokens t of qi.
  5. [Appendix A.1] Some dataset names in Table A1 (e.g., jinaai/ChartQA, jinaai/CharXiv-en) appear to be re-hosted versions of existing datasets; please state explicitly whether the splits differ from the originals.

Circularity Check

0 steps flagged · score 1.0 of 10

No derivation-level circularity: the model claims are empirical and independently benchmarked; the author-built Jina-VDR benchmark raises a contamination concern but no equation-level reduction is exhibited.

full rationale

This is an empirical systems paper rather than a derivation chain. The training losses (InfoNCE, CoSENT, matryoshka loss) and the LoRA adapter design contain no fitted constants that are later renamed as predictions, and no equation in Sections 4–5 defines a benchmark result as a consequence of a training input. The headline state-of-the-art claim on Jina-VDR is a measured evaluation outcome, not a derived identity. Furthermore, the paper is largely grounded in external benchmarks such as ViDoRe, CLIP Benchmark, MTEB, MMTEB, CoIR, and LongEmbed, so the central claims do not rest solely on self-citation. The paper does cite its own prior work (jina-embeddings-v3, jina-clip-v2) for methodological choices and data filtering, but those citations are not load-bearing for the main empirical conclusions. The most serious concern is that Jina-VDR is an author-constructed benchmark and Section 5.2.1 reports using hard negatives from “VDR multilingual” and other curated multimodal datasets during training, creating a plausible risk of overlap between training data and Jina-VDR test images. However, the paper does not exhibit an equation-level identity between training data and the evaluated score, and the analysis rules require quoting a specific reduction before a circularity step can be recorded. Therefore no circular step is listed; the score of 1 reflects a minor self-benchmark and self-citation burden rather than a demonstrated circular derivation.

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

This is an empirical systems paper without a mathematical derivation. The free parameters are training hyperparameters whose values are partly unreported. The axioms are domain assumptions about the pretrained backbone, negative sampling, benchmark validity, and pooling. No new physical or theoretical entities are introduced.

free parameters (4)
  • Contrastive temperature tau = 0.02 for the code adapter; unspecified for other adapters
    The temperature in InfoNCE and CoSENT losses controls gradient sharpness. Section 5.2.3 states 0.02 for triplet training, but Section 5.1 does not report the temperature used in pair training.
  • Joint loss weights w1 through w6 = not reported
    Equation 6 weights six loss terms balancing dense versus late interaction and text versus multimodal batches. No values or tuning procedure are reported.
  • Matryoshka loss weight and dimension schedule = not reported
    Matryoshka Representation Learning requires a dimension schedule or auxiliary loss weight to train truncatable single-vector embeddings. Section 5 states the loss is applied but gives no schedule.
  • LoRA rank and alpha = 60 million parameters per adapter
    The paper reports the parameter count but not the rank, alpha, or target modules. These are design choices that affect adapter capacity and are not independently justified.
assumptions (4)
  • domain assumption Qwen2.5-VL-3B-Instruct provides sufficiently strong pretrained multilingual text and vision representations for embedding fine-tuning.
    Section 4 initializes from this backbone and freezes it during training, so the model's ceiling depends on the backbone's pretrained quality.
  • domain assumption In-batch and mined hard negatives in contrastive training produce a representation that transfers to real retrieval.
    Section 5.1 and 5.2.1 rely on InfoNCE with in-batch negatives and mined hard negatives; if these negatives are too easy, the learned similarity may not reflect real query distributions.
  • domain assumption Jina-VDR benchmark queries and relevance labels measure genuine retrieval quality.
    Section 6 constructs queries from humans, templates, and generative models. The validity of the benchmark as a measure of real-world visual retrieval is assumed rather than demonstrated.
  • domain assumption Mean pooling of final-layer token embeddings preserves enough semantic information for single-vector retrieval.
    Section 4.2 uses mean pooling for single-vector output without an analysis of whether pooling discards retrieval-relevant detail.

how reviews work

0 comments
Cite this review

Pith. "Pith review of jina-embeddings-v4: Universal Embeddings for Multimodal Multilingual Retrieval." pith.science (2026). https://pith.science/paper/2KSABC4P

@misc{pith2026250618902,
  author       = {Pith},
  title        = {Pith review of: jina-embeddings-v4: Universal Embeddings for Multimodal Multilingual Retrieval},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2KSABC4P}},
  note         = {Machine review of arXiv:2506.18902}
}
read the original abstract

We introduce jina-embeddings-v4, a 3.8 billion parameter multimodal embedding model that unifies text and image representations through a novel architecture supporting both single-vector and multi-vector embeddings in the late interaction style. The model incorporates task-specific Low-Rank Adaptation (LoRA) adapters to optimize performance across diverse retrieval scenarios, including query-document retrieval, semantic text similarity, and code search. Comprehensive evaluations demonstrate that jina-embeddings-v4 achieves state-of-the-art performance on both single-modal and cross-modal retrieval tasks, with particular strength in processing visually rich content such as tables, charts, diagrams, and mixed-media formats. To facilitate evaluation of this capability, we also introduce Jina-VDR, a novel benchmark specifically designed for visually rich image retrieval.

Figures

Figures reproduced from arXiv: 2506.18902 by the authors.

Figure 1
Figure 1. Architecture of jina-embeddings-v4. The model employs a unified LM built on the Qwen2.5-VL-3B￾Instruct backbone (3.8B parameters). Text and image inputs are processed through a shared pathway: images are first converted to token sequences via a vision encoder, then both modalities are jointly processed by the language model decoder with contextual attention layers. Three task-specific LoRA adapters (60M parameters e… view at source ↗
Figure 2
Figure 2. Distribution of the cosine similarities of [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Distribution of the cosine similarities of [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 8 Pith papers

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

  1. LMEB: Long-horizon Memory Embedding Benchmark

    cs.CL 2026-03 unverdicted novelty 7.0 of 10

    LMEB is a new benchmark that evaluates embedding models on long-horizon memory retrieval and shows this skill is largely orthogonal to traditional passage-retrieval performance.

  2. Rethinking Text-Based Image Retrieval in Specific Domain

    cs.CV 2026-08 conditional novelty 6.0 of 10

    A new multi-match surveillance TBIR benchmark and a fine-tuning framework with cross-modal soft labels and intra-modal distillation improve mAP@20 by 7.8 points over standard contrastive tuning.

  3. Do Current Retrievers Cover All the Evidence? A Controlled Study of Conjunctive Cross-Page Retrieval

    cs.CV 2026-07 conditional novelty 6.0 of 10

    On conjunctive cross-page retrieval, strong hybrids find all-condition golds for ~81% of queries but rank them before natural subset matches only ~36% of the time.

  4. Learn to Pool: Lightweight Fine-Tuning for Flexible Multi-Vector Compression

    cs.IR 2026-07 conditional novelty 5.0 of 10

    Lightweight pooling-aware fine-tuning with k-means on a single dataset enables up to 83% vector compression in ColBERT models with no retrieval accuracy loss and positive cross-dataset transfer.

  5. Advancing Subsurface Discovery and Geothermal Monitoring with an Agentic Artificial Intelligence Framework

    physics.geo-ph 2025-11 reject novelty 5.0 of 10

    GAIA couples an LLM planning agent with retrieval-augmented search and standard seismological tools to automate geothermal monitoring, but no quantitative evaluation is reported.

  6. Efficient Code Embeddings from Code Generation Models

    cs.CL 2025-08 conditional novelty 5.0 of 10

    jina-code-embeddings, 0.5B and 1.5B models built from Qwen2.5-Coder with last-token pooling, achieve strong code retrieval scores competitive with far larger embedding models.

  7. VulCoCo: A Simple Yet Effective Method for Detecting Vulnerable Code Clones

    cs.SE 2025-07 conditional novelty 5.0 of 10

    VulCoCo retrieves candidate code clones with embeddings and validates them with an LLM, outperforming prior vulnerable-clone detectors on a new synthetic benchmark and finding real-world clones that led to 15 CVEs.

  8. ColGraphRAG: Late-Interaction Evidence Retrieval for Multimodal GraphRAG

    cs.AI 2026-05 conditional novelty 4.0 of 10

    Swapping pooled visual similarity for late-interaction MaxSim in graph-grounded multimodal QA is reported to improve graph-linked image retrieval and QA point estimates on MultimodalQA.

Reference graph

Works this paper leans on

30 extracted references · 12 canonical work pages · cited by 8 Pith papers

  1. [3]

    jina-embeddings-v3: Multilingual Embeddings With Task LoRA.arXiv preprint arXiv:2409.10173,

    Saba Sturua, Isabelle Mohr, Mohammad Kalim Akram, Michael Günther, Bo Wang, Markus Krimmel, Feng Wang, Georgios Mastrapas, Andreas Koukounas, Nan Wang, et al. jina-embeddings-v3: Multilingual Embeddings With Task LoRA.arXiv preprint arXiv:2409.10173,

    arXiv

  2. [6]

    Xueguang Ma, Sheng-Chieh Lin, Minghan Li, Wenhu Chen, and Jimmy Lin

    https://blog.voyageai.com/2024/04/15/domain- specific-embeddings-and-retrieval-legal-edition- voyage-law-2/. Xueguang Ma, Sheng-Chieh Lin, Minghan Li, Wenhu Chen, and Jimmy Lin. Unifying Multimodal Retrieval via Document Screenshot Embedding. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 6492–6505,

    work page 2024

  3. [9]

    Jina Embeddings 2: 8192- Token General-Purpose Text Embeddings for Long Documents.arXiv preprint arXiv:2310.19923,

    Michael Günther, Jackmin Ong, Isabelle Mohr, Alaed- dine Abdessalem, Tanguy Abel, Mohammad Kalim Akram, Susana Guzman, Georgios Mastrapas, Saba Sturua, Bo Wang, et al. Jina Embeddings 2: 8192- Token General-Purpose Text Embeddings for Long Documents.arXiv preprint arXiv:2310.19923,

    arXiv

  4. [10]

    Qwen2.5-VL Technical Report.arXiv preprint arXiv:2502.13923,

    Shuai Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, Sibo Song, Kai Dang, Peng Wang, Shijie Wang, Jun Tang, Humen Zhong, Yuanzhi Zhu, Mingkun Yang, Zhaohai Li, Jianqiang Wan, Pengfei Wang, Wei Ding, Zheren Fu, Yiheng Xu, Jiabo Ye, Xi Zhang, Tianbao Xie, Zesen Cheng, Hang Zhang, Zhibo Yang, Haiyang Xu, and Junyang Lin. Qwen2.5-VL Technical Report.ar...

    arXiv

  5. [11]

    E5-V: Universal Embeddings with Multimodal Large Language Models.arXiv preprint arXiv:2407.12580,

    Ting Jiang, Minghui Song, Zihan Zhang, Haizhen Huang, Weiwei Deng, Feng Sun, Qi Zhang, Deqing Wang, and Fuzhen Zhuang. E5-V: Universal Embeddings with Multimodal Large Language Models.arXiv preprint arXiv:2407.12580,

    arXiv

  6. [12]

    GME: Improving Universal Multimodal Retrieval by Multimodal LLMs.arXiv preprint arXiv:2412.16855,

    Xin Zhang, Yanzhao Zhang, Wen Xie, Mingxin Li, Ziqi Dai, Dingkun Long, Pengjun Xie, Meishan Zhang, Wenjie Li, and Min Zhang. GME: Improving Universal Multimodal Retrieval by Multimodal LLMs.arXiv preprint arXiv:2412.16855,

    arXiv

  7. [16]

    Semeval-2012 task 6: A pilot on semantic textual similarity

    Eneko Agirre, Daniel Cer, Mona Diab, and Aitor Gonzalez-Agirre. Semeval-2012 task 6: A pilot on semantic textual similarity. InSEM 2012: The First Joint Conference on Lexical and Computational Semantics–V olume 1: Proceedings of the main confer- ence and the shared task, and V olume 2: Proceedings of the Sixth International Workshop on Semantic Evaluation...

    work page 2012

  8. [18]

    OpenCodeInterpreter: Integrating Code Generation with Execution and Refinement

    Tianyu Zheng, Ge Zhang, Tianhao Shen, Xueling Liu, Bill Yuchen Lin, Jie Fu, Wenhu Chen, and Xiang Yue. OpenCodeInterpreter: Integrating Code Generation with Execution and Refinement. In Lun-Wei Ku, Andre Martins, and Vivek Srikumar, editors,Findings of the Association for Computational Linguistics: ACL 2024, pages 12834–12859. Association for Computationa...

    work page 2024

Show all 30 references
  1. [19]

    Cornstack: High-quality contrastive data for better code retrieval and reranking.arXiv preprint arXiv:2412.01007,

    Tarun Suresh, Revanth Gangi Reddy, Yifei Xu, Zach Nussbaum, Andriy Mulyar, Brandon Duderstadt, and Heng Ji. Cornstack: High-quality contrastive data for better code retrieval and reranking.arXiv preprint arXiv:2412.01007,

    arXiv

  2. [22]

    Available at: https://jina.ai/news/re-search- order-2024-yearbook-of-search-foundation- advances/

    Hardcover with spot UV coating; includes complimentary digital copy. Available at: https://jina.ai/news/re-search- order-2024-yearbook-of-search-foundation- advances/. Niigata-shi Kank ¯o Kokusaik ¯ory¯ubu Kank ¯o Su- ishinka.Niigata City Ramen Guidebook. City of Niigata, Niig...

    2024

  3. [23]

    PDF, approximately 28 MB

    URL https://www.city.niigata.lg.jp/kanko/kanko/ oshirase/ramen.files/guidebook.pdf. PDF, approximately 28 MB. Shanghai Municipal People’s Government Urban Planning and Land Resource Administration Bureau. Shanghai Master Plan 2017–2035: Striving for the Excellent Global City. ...

    2017

  4. [24]

    Public Reading edition; government-issued planning document

    URL https://www.shanghai.gov.cn/newshanghai/ xxgkfj/2035004.pdf. Public Reading edition; government-issued planning document. Sara Ghaboura, Ahmed Heakl, Omkar Thawakar, Ali Alharthi, Ines Riahi, Abduljalil Saif, Jorma Laaksonen, Fahad S. Khan, Salman Khan, and Rao M. Anwer. C...

  5. [25]

    Cyrile Delestre,

    URL https://arxiv.org/abs/2410.18976. Cyrile Delestre,

    arXiv

  6. [26]

    Improving text embeddings with large language models.arXiv preprint arXiv:2401.00368,

    Liang Wang, Nan Yang, Xiaolong Huang, Linjun Yang, Rangan Majumder, and Furu Wei. Improving text embeddings with large language models.arXiv preprint arXiv:2401.00368,

    arXiv

  7. [27]

    LongEmbed: Extending Embedding Models for Long Context Retrieval

    Dawei Zhu, Liang Wang, Nan Yang, Yifan Song, Wenhao Wu, Furu Wei, and Sujian Li. LongEmbed: Extending Embedding Models for Long Context Retrieval. InProceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 802–816,

    2024

  8. [28]

    Thapliyal, Jordi Pont Tuset, Xi Chen, and Radu Soricut

    Ashish V . Thapliyal, Jordi Pont Tuset, Xi Chen, and Radu Soricut. Crossmodal-3600: A massively multilingual multimodal evaluation dataset. InProceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 715–729,

    2022

  9. [29]

    Mind the gap: Under- standing the modality gap in multi-modal contrastive representation learning.Advances in Neural Infor- mation Processing Systems, 35:17612–17625, 2022b

    Victor Weixin Liang, Yuhui Zhang, Yongchan Kwon, Ser- ena Yeung, and James Y Zou. Mind the gap: Under- standing the modality gap in multi-modal contrastive representation learning.Advances in Neural Infor- mation Processing Systems, 35:17612–17625, 2022b. Simon Schrodi, David ...

    arXiv

  10. [30]

    Lawrence Zitnick

    Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Dollár, and C. Lawrence Zitnick. Microsoft coco: Common objects in context. In David Fleet, Tomas Pajdla, Bernt Schiele, and Tinne Tuytelaars, editors,Com- puter Vision – ECCV 2014, pag...

    2014

  11. [97]

    16 91.66 91.48 94.92 94.48 65.35 jina-embeddings-v4 (late) 90.17 88.95 59.98 93.57 92.35 99.26 96.76 96.95 98.39 95.13 80.34 Tasks:Avg: Mean nDCG@10% over all tasks, AQA: ArxivQA, Shift: Shift Project, DVQA: DocVQA, InfoVQA: InfographicVQA, AI: Artificial Intelligence, Gov: Go...

  12. [2014]

    URL http://www.lrec-conf.org/proceedings/ lrec2014/pdf/363_Paper.pdf

    European Language Resources Association (ELRA). URL http://www.lrec-conf.org/proceedings/ lrec2014/pdf/363_Paper.pdf. Hamel Husain, Ho-Hsiang Wu, Tiferet Gazit, Miltiadis Allamanis, and Marc Brockschmidt. CodeSearchNet Challenge: Evaluating the State of Semantic Code Search.ar...

    1909 arXiv

  13. [2015]

    Towards General Text Embeddings with Multi-stage Contrastive Learning.arXiv preprint arXiv:2308.03281,

    Zehan Li, Xin Zhang, Yanzhao Zhang, Dingkun Long, Pengjun Xie, and Meishan Zhang. Towards General Text Embeddings with Multi-stage Contrastive Learning.arXiv preprint arXiv:2308.03281,

    arXiv

  14. [2017]

    Jina AI.Re·Search: Order 2024 Yearbook of Search F oundation Advances

    doi:10.1109/CVPR.2017.571. Jina AI.Re·Search: Order 2024 Yearbook of Search F oundation Advances. Jina AI, Sunnyvale, CA & Berlin, Germany, December 16

    2017 doi

  15. [2018]

    Geoffrey Hinton, Oriol Vinyals, and Jeff Dean

    URL http://arxiv.org/abs/1807.03748. Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. Distilling the Knowledge in a Neural Network.arXiv preprint arXiv:1503.02531,

    arXiv

  16. [2019]

    Text Embeddings by Weakly- Supervised Contrastive Pre-training.arXiv preprint arXiv:2212.03533,

    Liang Wang, Nan Yang, Xiaolong Huang, Binxing Jiao, Linjun Yang, Daxin Jiang, Rangan Majumder, and Furu Wei. Text Embeddings by Weakly- Supervised Contrastive Pre-training.arXiv preprint arXiv:2212.03533,

    arXiv

  17. [2020]

    Sentence-BERT: Sen- tence Embeddings using Siamese BERT-Networks

    Nils Reimers and Iryna Gurevych. Sentence-BERT: Sen- tence Embeddings using Siamese BERT-Networks. InProceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP),...

    2019

  18. [2021]

    Jina CLIP: Your CLIP Model Is Also Your Text Retriever.arXiv preprint arXiv:2405.20204,

    Andreas Koukounas, Georgios Mastrapas, Michael Gün- ther, Bo Wang, Scott Martens, Isabelle Mohr, Saba Sturua, Mohammad Kalim Akram, Joan Fontanals Martínez, Saahil Ognawala, et al. Jina CLIP: Your CLIP Model Is Also Your Text Retriever.arXiv preprint arXiv:2405.20204,

    arXiv

  19. [2022]

    Deep learning based visually rich document content understanding: A survey.arXiv preprint arXiv:2408.01287,

    Yihao Ding, Soyeon Caren Han, Jean Lee, and Eduard Hovy. Deep learning based visually rich document content understanding: A survey.arXiv preprint arXiv:2408.01287,

    arXiv

  20. [2023]

    Zirui Wang, Mengzhou Xia, Luxi He, Howard Chen, Yitao Liu, Richard Zhu, Kaiqu Liang, Xindi Wu, Haotian Liu, Sadhika Malladi, Alexis Chevalier, Sanjeev Arora, and Danqi Chen

    URL https://arxiv.org/abs/2304.09660. Zirui Wang, Mengzhou Xia, Luxi He, Howard Chen, Yitao Liu, Richard Zhu, Kaiqu Liang, Xindi Wu, Haotian Liu, Sadhika Malladi, Alexis Chevalier, Sanjeev Arora, and Danqi Chen. Charxiv: Charting gaps in realistic chart understanding in multim...

    arXiv

  21. [2024]

    CodeX- Embed: A Generalist Embedding Model Family for Multilingual and Multi-task Code Retrieval.arXiv preprint arXiv:2411.12644,

    Ye Liu, Rui Meng, Shafiq Joty, Silvio Savarese, Caiming Xiong, Yingbo Zhou, and Semih Yavuz. CodeX- Embed: A Generalist Embedding Model Family for Multilingual and Multi-task Code Retrieval.arXiv preprint arXiv:2411.12644,

    arXiv

  22. [2025]

    MIEB: Massive Image Embedding Benchmark.arXiv preprint arXiv:2504.10471,

    Chenghao Xiao, Isaac Chung, Imene Kerboua, Jamie Stir- ling, Xin Zhang, Márton Kardos, Roman Solomatin, Noura Al Moubayed, Kenneth Enevoldsen, and Niklas Muennighoff. MIEB: Massive Image Embedding Benchmark.arXiv preprint arXiv:2504.10471,

    arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.