Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

Towards Storage-Efficient Visual Document Retrieval: An Empirical Study on Reducing Patch-Level Embeddings

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Token merging, not pruning, compresses ColPali/ColQwen2 indexes 9x-35x while retaining roughly 95-98% of retrieval performance.

desk verdict Solid empirical study on compressing ColPali-style patch embeddings; merging at the last layer with fine-tuning works, and the negative result on pruning is the most interesting part. read the letter →

arxiv 2506.04997 v1 pith:RY4PNJH3 submitted 2025-06-05 cs.IR cs.CL

Yubo Ma , Jinsong Li , Yuhang Zang , Xiaobao Wu , Xiaoyi Dong , Pan Zhang , Yuhang Cao , Haodong Duan
show 3 more authors
Jiaqi Wang Yixin Cao Aixin Sun
This is my paper
classification cs.IRcs.CL
keywords visualdocumentretrievalColPaliQwen2tokenmergingpruningpatch-levelembeddingsstorageefficiencylateinteraction
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper asks whether the large memory footprint of ColPali/ColQwen2—one 128-dimensional embedding per image patch per page—can be cut without sacrificing retrieval quality. It finds that token pruning cannot: the patches a query activates are almost disjoint across different queries, so attention- and score-based pruning underperform random dropping and collapse under aggressive ratios. Token merging works because page patches are highly redundant and clusterable. The proposed Light-ColPali/ColQwen2 clusters patch embeddings by cosine similarity at the final projection layer, averages each cluster into one prototype, and fine-tunes the retriever on the merged representations. It retains 98.2% of NDCG@5 with 11.8% of the original embedding memory, and 94.6% at 2.8% memory.

What carries the argument

The central object is the MaxSim relevance score $s(q,p)=\sum_j \max_i e_i^p{}^\top e_j^q$, which sums, over query tokens, the best matching patch embedding per page. The carrying mechanism is semantic clustering: hierarchical agglomerative clustering of the patch embeddings' cosine similarities, applied after the final projection layer where dimension is lowest and therefore clustering is most effective, with each cluster replaced by its mean vector. Fine-tuning the retriever with these merged embeddings during both training and inference is what recovers most of the performance lost by training-free merging; the merging factor $N_p/N'_p$ sets the storage reduction.

What would settle it

Compare original ColQwen2 and Light-ColQwen2 at merging factor 9 on a held-out set of real, non-synthetic queries, focusing on text-dense pages such as DocVQA. If a query subset shows a large drop in retrieval of the ground-truth page relative to the 98.2% average—equivalently, if the top-activating patches for those queries are consistently merged away—the query-independent redundancy assumption fails.

Watch

Extended reading notes

Core claim

The central claim is that the stored patch-level embeddings of ColPali/ColQwen2 are substantially redundant, and that redundancy can be exploited by merging rather than pruning. The evidence has three parts: response-potential distributions are clustered, with on average 36.9 patches per page above normalized 0.9 and 14.2 above 0.95; activated patches are query-dependent, so pruning must guess which patches to keep and random dropping wins only because it does not systematically delete whole clusters; and semantic clustering at the post-projector location, followed by fine-tuning, preserves MaxSim-based relevance. Across nine ViDoRE, VisRAG, and MMLongBench-Doc datasets, Light-ColPali/ColQwen2 keeps 99.0% of NDCG@5 at merging factor 4, 98.2% at factor 9, and 94.6% at factor 49, at which point its memory is comparable to single-vector DSE baselines.

Load-bearing premise

Patch-level embeddings are redundant in a query-independent way, so clustering them by cosine similarity and averaging within clusters preserves MaxSim relevance for arbitrary unseen queries; the paper's redundancy evidence uses only five synthetic queries per page, and real queries could activate different patch subsets.

Editorial extensions

If this is right

  • Light-ColQwen2 keeps 99.0% of NDCG@5 at a merging factor of 4 (25.5% memory) and 98.2% at factor 9 (11.8% memory), so deployment can choose a Pareto point along the reduction curve.
  • At merging factor 49, the compressed retriever stores 1.8x (Qwen2) or 0.9x (PaliGemma) the memory of a single-embedding DSE retriever while still beating DSE in absolute NDCG@5.
  • Training-free semantic clustering at factor 9 already retains roughly 97.5% average performance, meaning the merging recipe works even without fine-tuning existing indexes.
  • Token pruning never reaches competitive retention: at a 0.9-0.95 pruning ratio the best strategy keeps only 58-88% of the original score, so pruning is not a viable route to order-of-magnitude reduction.
  • Fine-tuning recovers 61% of the performance drop at merging factor 25 and 67% at factor 49, so the marginal cost of extreme compression is mostly recoverable with training.

Reading between the lines

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

  • The query-independent redundancy assumption implies a page-adaptive merging factor: pages with high information density should merge less, and a cheap density estimate could pick the factor per page at index time; the paper notes adaptivity as open future work.
  • The same late-stage semantic-clustering recipe is a candidate for text-based multi-vector retrievers, where word embeddings may be at least as clusterable as visual patches.
  • Because the merging module runs offline on stored embeddings, it can be applied post hoc to an already-built ColPali/ColQwen2 index without retraining; the gap to the fine-tuned version then measures how much compression is available for free.
  • The paper's redundancy evidence rests on five synthetic queries per page; a stress test with human queries across new domains would show whether the 94-98% retention transfers outside the nine benchmark datasets.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper studies storage reduction for the patch-level embeddings produced by ColPali/ColQwen2 visual document retrievers. It compares two token-reduction families: pruning and merging. On pruning, it reports that random pruning outperforms score- and attention-based strategies, but all pruning methods degrade substantially at aggressive ratios. On merging, it evaluates three dimensions—merging approach (1D/2D spatial pooling vs. semantic clustering), fine-tuning applicability, and merging location (pre-encoder to post-projector)—and finds that late-stage semantic clustering combined with fine-tuning works best. The resulting method, Light-ColPali/ColQwen2, is reported to retain 98.2% of NDCG@5 at 11.8% of original memory and 94.6% at 2.8% memory, averaged over nine datasets from ViDoRE, VisRAG, and MMLongBench-Doc.

Significance. If the reported retention figures hold, this is a practical 9x–35x compression of document embedding storage with small retrieval loss, and the paper provides a systematic comparison of pruning versus merging under the VDR setting. The strengths are the broad evaluation across nine datasets, the combination of training-free and fine-tuned results, and the honest reporting of time costs. The main weakness is that the central assumption—that patch embeddings are redundant in a query-independent way—is supported only by an analysis based on five synthetic queries per page, and no experiment tests transfer to held-out, topic-disjoint queries. The lack of error bars further weakens the precision of the headline numbers.

major comments (4)
  1. [Section 4.3 / Appendix A.2] The central claim that merging via semantic clustering preserves retrieval for arbitrary unseen queries is not supported by the evidence presented. For a merged cluster C_k, the score for a query token e_q is max_k (avg_{i in C_k} e_i)^T e_q, which is bounded above by max_i e_i^T e_q; equality holds only if all members of the winning cluster have identical dot products with e_q. The redundancy analysis in Section 4.3 shows that about 14.2 patches per page have normalized response above 0.95 for a given query, but this does not establish that cosine-similarity clusters are stable across queries. In fact, Figure 3(a) shows that activated patches for two synthetic queries are nearly disjoint, and the synthetic queries themselves are generated by Qwen2-VL-7B. Fine-tuning in Section 5.2 is performed on the ColPali training distribution, so it may compensate for information loss only on query patterns similar to training. No experiment measures retention on a held-out, topic-disjoint query distribution. Please add such an evaluation or provide a direct analysis of within-cluster response-potential variance across diverse query sets; without this, the 98.2% and 94.6% figures are not yet established as a general property of merging.
  2. [Tables 2 and 4] All reported NDCG@5 values are single-run point estimates with no error bars or significance tests. The differences between merging factors (e.g., 99.0% at factor 4, 98.2% at factor 9, 96.3% at factor 25 for Light-ColQwen2) are small, and without variance estimates it is unclear whether these differences are meaningful. Please report means and standard deviations over at least three independent fine-tuning runs, or conduct paired significance tests (e.g., bootstrap or paired t-test) for the key comparisons against ColQwen2 and ColPali.
  3. [Section 5.2] The fine-tuning procedure is underspecified with respect to the clustering module. It is not stated whether the cluster assignments for the merged document embeddings are recomputed at each training step as the model weights change, or fixed using the initial model's embeddings. This matters because if clusters are fixed, the model is trained against a representation that becomes stale as embeddings drift; if recomputed, the training objective changes during optimization. Please clarify and justify the choice, as it directly affects the interpretation of the fine-tuning gains in Figure 6 and the reproducibility of the method.
  4. [Section 4.2] The conclusion that pruning is 'inherently unsuitable' for VDR is based on experiments on only two datasets (DocVQA and InfoVQA) and on synthetic queries generated by Qwen2-VL-7B. While Table 2 extends random pruning to nine datasets, the comparison of the three pruning strategies and the analysis of query-dependent activation are limited to these two datasets. This strong negative claim should either be supported on the full benchmark suite or tempered to a claim about the tested conditions.
minor comments (6)
  1. [Figure 3(a)] The definition of Overlap@R is not given in the caption or text; please state how overlap is computed and what the dashed diagonal represents.
  2. [Section 5.1] Please specify the linkage criterion (e.g., average, Ward) and the distance metric used in the hierarchical clustering procedure.
  3. [Table 2] The 'Average' column appears to report mean NDCG@5, but the caption also mentions relative performance; please clarify what the average column contains and how the relative performance percentages are derived.
  4. [Table 2 / Section 6] The 'ColPali+Pruning' baseline uses random pruning, but this is not stated in the table or its caption; please state it explicitly.
  5. [Abstract / entire manuscript] There are several spacing and formatting typos, such as 'to-ken pruningandtoken merging' in the abstract and 'ColPali/-ColQwen2' in the introduction; please proofread.
  6. [Limitations] The Limitations section does not mention the potential issue that the merging strategy's effectiveness may depend on query distribution; please add a discussion of this limitation, particularly in light of the synthetic-query-based analysis.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the compression claims are empirical and benchmarked against external datasets; no central result reduces to its own inputs.

full rationale

No circular step is present. The paper's main claim is that Light-ColPali/ColQwen2 preserves 98.2% of NDCG@5 at 11.8% memory and 94.6% at 2.8% memory; these numbers are measured on external benchmarks (ViDoRE, VisRAG, MMLongBench-Doc) rather than derived from the method's construction. The merging recipe (semantic clustering, late-stage merging, fine-tuning) is selected by comparing configurations on held-out test sets, and the fine-tuning uses training queries while evaluation uses benchmark queries. The synthesized queries in Section 4.1 and Appendix A.1 are used only to analyze pruning behavior, not to fit or define the final retrieval model, so no fitted input is renamed as a prediction. References to the authors' own MMLongBench-Doc work are dataset citations, not load-bearing theoretical premises. No uniqueness theorem or prior-work ansatz is used to force the design choice. The main validity risk is the limited evidence for query-independent patch redundancy (five synthetic queries per page), but that is a generalization concern, not circularity.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The central claim rests on the domain assumption that query-agnostic cluster averaging preserves retrieval-relevant information, plus a practical assumption about the synthetic queries used in the pruning analysis. No free parameters are fitted to the test data; merging factors are swept and reported rather than tuned to a target. No new physical or conceptual entities are introduced.

assumptions (3)
  • domain assumption Cosine similarity among patch embeddings is a valid measure of semantic redundancy for retrieval.
    Used to justify semantic clustering in Section 5.1; no evidence that embedding-space proximity aligns with answer content for unseen queries.
  • domain assumption Five synthesized queries per page approximate the distribution of real queries over patch activations.
    Used in Section 4.3 and Appendix A.1 to conclude that pruning is inherently unsuitable; real query distributions may differ.
  • domain assumption Gradients can be propagated through the non-differentiable hierarchical clustering operation during fine-tuning.
    Section 5.2 uses merged embeddings during training but does not specify the differentiation mechanism.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Storage-Efficient Visual Document Retrieval: An Empirical Study on Reducing Patch-Level Embeddings." pith.science (2026). https://pith.science/paper/RY4PNJH3

@misc{pith2026250604997,
  author       = {Pith},
  title        = {Pith review of: Towards Storage-Efficient Visual Document Retrieval: An Empirical Study on Reducing Patch-Level Embeddings},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RY4PNJH3}},
  note         = {Machine review of arXiv:2506.04997}
}
read the original abstract

Despite the strong performance of ColPali/ColQwen2 in Visualized Document Retrieval (VDR), it encodes each page into multiple patch-level embeddings and leads to excessive memory usage. This empirical study investigates methods to reduce patch embeddings per page at minimum performance degradation. We evaluate two token-reduction strategies: token pruning and token merging. Regarding token pruning, we surprisingly observe that a simple random strategy outperforms other sophisticated pruning methods, though still far from satisfactory. Further analysis reveals that pruning is inherently unsuitable for VDR as it requires removing certain page embeddings without query-specific information. Turning to token merging (more suitable for VDR), we search for the optimal combinations of merging strategy across three dimensions and develop Light-ColPali/ColQwen2. It maintains 98.2% of retrieval performance with only 11.8% of original memory usage, and preserves 94.6% effectiveness at 2.8% memory footprint. We expect our empirical findings and resulting Light-ColPali/ColQwen2 offer valuable insights and establish a competitive baseline for future research towards efficient VDR.

Figures

Figures reproduced from arXiv: 2506.04997 by the authors.

Figure 1
Figure 1. Top: The relative memory consumptions for embedding storage of different VDRs. Our simple yet effective approach, Light-ColPali/ColQwen2, retains most of the performance but with significantly reduced memory cost. Bottom: The diagram of VDR equipped with ColPali/ColQwen2 retriever. It encodes each page into Np patch-level embeddings and thus incurs pro￾hibitive memory cost. potential across various applications (Cho… view at source ↗
Figure 2
Figure 2. Retrieval performance v.s. pruning ratio [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The triggered patches of the identical page [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Retrieval performance v.s. merging factor [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5 [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Training-free v.s. fine-tuning retriever with the [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Do All Visual Tokens Matter Equally? Object-Evidence Preserving Token Merging for Vision-Language Retrieval

    cs.IR 2026-07 conditional novelty 6.0 of 10

    Object-aware soft merging of post-projector visual tokens preserves MaxSim-selectable evidence, yielding >93% token reduction and higher R@1 than full ColPali on Flickr30K and MSCOCO.

Reference graph

Works this paper leans on

36 extracted references · 11 canonical work pages · cited by 1 Pith paper

  1. [1]

    Lucas Beyer, Andreas Steiner, André Susano Pinto, Alexander Kolesnikov, Xiao Wang, Daniel Salz, Maxim Neumann, Ibrahim Alabdulmohsin, Michael Tschannen, Emanuele Bugliarello, Thomas Unterthiner, Daniel Keysers, Skanda Koppula, Fangyu Liu, Adam Grycner, Alexey Gritsenko, Neil Houlsby, Manoj Kumar, Keran Rong, Julian Eisenschlos, Rishabh Kabra, Matthias Bau...

    arXiv 2024

  2. [2]

    Daniel Bolya, Cheng-Yang Fu, Xiaoliang Dai, Peizhao Zhang, Christoph Feichtenhofer, and Judy Hoffman. 2023. https://openreview.net/forum?id=JroZRaRw7Eu Token merging: Your vit but faster . In The Eleventh International Conference on Learning Representations

    work page 2023

  3. [3]

    Rossi, Changyou Chen, and Tong Sun

    Jian Chen, Ruiyi Zhang, Yufan Zhou, Tong Yu, Franck Dernoncourt, Jiuxiang Gu, Ryan A. Rossi, Changyou Chen, and Tong Sun. 2025. https://openreview.net/forum?id=FDaHjwInXO Lo RA -contextualizing adaptation of large multimodal models for multi-page document understanding . In The Thirteenth International Conference on Learning Representations

    work page 2025

  4. [4]

    Liang Chen, Haozhe Zhao, Tianyu Liu, Shuai Bai, Junyang Lin, Chang Zhou, and Baobao Chang. 2024. https://api.semanticscholar.org/CorpusID:268358224 An image is worth 1/2 tokens after layer 2: Plug-and-play inference acceleration for large vision-language models . In European Conference on Computer Vision

    work page 2024

  5. [5]

    Jaemin Cho, Debanjan Mahata, Ozan Irsoy, Yujie He, and Mohit Bansal. 2024. https://arxiv.org/abs/2411.04952 M3docrag: Multi-modal retrieval is what you need for multi-page multi-document understanding . Preprint, arXiv:2411.04952

    arXiv 2024

  6. [6]

    Benjamin Clavié, Antoine Chaffin, and Griffin Adams. 2024. https://arxiv.org/abs/2409.14683 Reducing the footprint of multi-vector retrieval with minimal performance impact via token pooling . Preprint, arXiv:2409.14683

    arXiv 2024

  7. [7]

    Manuel Faysse, Hugues Sibille, Tony Wu, Bilel Omrani, Gautier Viaud, CELINE HUDELOT, and Pierre Colombo. 2025. https://openreview.net/forum?id=ogjBpZ8uSi Colpali: Efficient document retrieval with vision language models . In The Thirteenth International Conference on Learning Representations

    work page 2025

  8. [8]

    Edward J Hu, yelong shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2022. https://openreview.net/forum?id=nZeVKeeFYf9 Lo RA : Low-rank adaptation of large language models . In International Conference on Learning Representations

    2022

Show all 36 references
  1. [9]

    Herve Jégou, Matthijs Douze, and Cordelia Schmid. 2011. https://doi.org/10.1109/TPAMI.2010.57 Product quantization for nearest neighbor search . IEEE Transactions on Pattern Analysis and Machine Intelligence, 33(1):117--128

    2011 doi

  2. [10]

    Vladimir Karpukhin, Barlas Oguz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih. 2020. https://doi.org/10.18653/v1/2020.emnlp-main.550 Dense passage retrieval for open-domain question answering . In Proceedings of the 2020 Conference on Empiric...

    2020 doi

  3. [11]

    Omar Khattab and Matei Zaharia. 2020. https://doi.org/10.1145/3397271.3401075 Colbert: Efficient and effective passage search via contextualized late interaction over bert . In Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Informatio...

    2020

  4. [12]

    Zhenglun Kong, Peiyan Dong, Xiaolong Ma, Xin Meng, Wei Niu, Mengshu Sun, Xuan Shen, Geng Yuan, Bin Ren, Hao Tang, et al. 2022. Spvit: Enabling faster vision transformers via latency-aware soft token pruning. In Computer Vision--ECCV 2022: 17th European Conference, Tel Aviv, Is...

    2022

  5. [13]

    Lei Li, Yuqi Wang, Runxin Xu, Peiyi Wang, Xiachong Feng, Lingpeng Kong, and Qi Liu. 2024 a . https://doi.org/10.18653/v1/2024.acl-long.775 Multimodal A r X iv: A dataset for improving scientific comprehension of large vision-language models . In Proceedings of the 62nd Annual ...

    2024 doi

  6. [14]

    Wentong Li, Yuqian Yuan, Jian Liu, Dongqi Tang, Song Wang, Jianke Zhu, and Lei Zhang. 2024 b . https://arxiv.org/abs/2407.02392 Tokenpacker: Efficient visual projector for multimodal llm . Preprint, arXiv:2407.02392

    2024 arXiv

  7. [15]

    Youwei Liang, Chongjian Ge, Zhan Tong, Yibing Song, Jue Wang, and Pengtao Xie. 2022. https://api.semanticscholar.org/CorpusID:246867285 Not all patches are what you need: Expediting vision transformers via token reorganizations . ArXiv, abs/2202.07800

    2022 arXiv

  8. [16]

    Xueguang Ma, Sheng-Chieh Lin, Minghan Li, Wenhu Chen, and Jimmy Lin. 2024 a . https://doi.org/10.18653/v1/2024.emnlp-main.373 Unifying multimodal retrieval via document screenshot embedding . In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Proces...

    2024 doi

  9. [17]

    Yubo Ma, Yuhang Zang, Liangyu Chen, Meiqi Chen, Yizhu Jiao, Xinze Li, Xinyuan Lu, Ziyu Liu, Yan Ma, Xiaoyi Dong, Pan Zhang, Liangming Pan, Yu-Gang Jiang, Jiaqi Wang, Yixin Cao, and Aixin Sun. 2024 b . https://openreview.net/forum?id=loJM1acwzf MMLONGBENCH - DOC : Benchmarking ...

    2024

  10. [18]

    Ahmed Masry, Do Xuan Long, Jia Qing Tan, Shafiq Joty, and Enamul Hoque. 2022. https://arxiv.org/abs/2203.10244 Chartqa: A benchmark for question answering about charts with visual and logical reasoning . Preprint, arXiv:2203.10244

    2022 arXiv

  11. [19]

    Minesh Mathew, Viraj Bagal, Rub \`e n P \'e rez Tito, Dimosthenis Karatzas, Ernest Valveny, and C.V. Jawahar. 2021. https://api.semanticscholar.org/CorpusID:233394125 Infographicvqa . 2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), pages 2582--2591

    2021

  12. [20]

    Manmatha, and C

    Minesh Mathew, Dimosthenis Karatzas, R. Manmatha, and C. V. Jawahar. 2020. https://api.semanticscholar.org/CorpusID:220280200 Docvqa: A dataset for vqa on document images . 2021 IEEE Winter Conference on Applications of Computer Vision (WACV), pages 2199--2208

    2020

  13. [21]

    Fionn Murtagh and Pedro Contreras. 2012. https://api.semanticscholar.org/CorpusID:18990050 Algorithms for hierarchical clustering: an overview . Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 2

    2012

  14. [22]

    Keshav Santhanam, Omar Khattab, Jon Saad-Falcon, Christopher Potts, and Matei Zaharia. 2022. https://doi.org/10.18653/v1/2022.naacl-main.272 C ol BERT v2: Effective and efficient retrieval via lightweight late interaction . In Proceedings of the 2022 Conference of the North Am...

    2022 doi

  15. [23]

    Yuzhang Shang, Mu Cai, Bingxin Xu, Yong Jae Lee, and Yan Yan. 2024. https://arxiv.org/abs/2403.15388 Llava-prumerge: Adaptive token reduction for efficient large multimodal models . Preprint, arXiv:2403.15388

    2024

  16. [24]

    Ryota Tanaka, Kyosuke Nishida, Kosuke Nishida, Taku Hasegawa, Itsumi Saito, and Kuniko Saito. 2023. Slidevqa: A dataset for document visual question answering on multiple images. In AAAI

    2023

  17. [25]

    Peng Wang, Shuai Bai, Sinan Tan, Shijie Wang, Zhihao Fan, Jinze Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, Yang Fan, Kai Dang, Mengfei Du, Xuancheng Ren, Rui Men, Dayiheng Liu, Chang Zhou, Jingren Zhou, and Junyang Lin. 2024. https://arxiv.org/abs/2409.12191 Qwen2-v...

    2024 arXiv

  18. [26]

    Tony Wu. 2024. https://x.com/tonywu_71/status/1826238532241010721 Token P ooling for C olpali

    2024

  19. [27]

    Xiaobao Wu. 2025. https://arxiv.org/abs/2505.02686 Sailing ai by the stars: A survey of learning from rewards in post-training and test-time scaling of large language models . arXiv preprint arXiv:2505.02686

    2025 arXiv

  20. [28]

    Xiaobao Wu, Liangming Pan, William Yang Wang, and Anh Tuan Luu. 2024 a . https://doi.org/10.18653/v1/2024.emnlp-main.843 AKEW : Assessing knowledge editing in the wild . In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 15118--151...

    2024 doi

  21. [29]

    Xiaobao Wu, Liangming Pan, Yuxi Xie, Ruiwen Zhou, Shuai Zhao, Yubo Ma, Mingzhe Du, Rui Mao, Anh Tuan Luu, and William Yang Wang. 2024 b . https://arxiv.org/pdf/2412.13670 AntiLeak-Bench : Preventing data contamination by automatically constructing benchmarks with updated real-...

    2024 arXiv

  22. [30]

    Long Xing, Qidong Huang, Xiaoyi Dong, Jiajie Lu, Pan Zhang, Yuhang Zang, Yuhang Cao, Conghui He, Jiaqi Wang, Feng Wu, and Dahua Lin. 2024. https://arxiv.org/abs/2410.17247 Pyramiddrop: Accelerating your large vision-language models via pyramid visual redundancy reduction . Pre...

    2024 arXiv

  23. [31]

    Senqiao Yang, Yukang Chen, Zhuotao Tian, Chengyao Wang, Jingyao Li, Bei Yu, and Jiaya Jia. 2024. https://arxiv.org/abs/2412.04467 Visionzip: Longer is better but not necessary in vision language models . Preprint, arXiv:2412.04467

    2024

  24. [32]

    Shi Yu, Chaoyue Tang, Bokai Xu, Junbo Cui, Junhao Ran, Yukun Yan, Zhenghao Liu, Shuo Wang, Xu Han, Zhiyuan Liu, and Maosong Sun. 2024. https://arxiv.org/abs/2410.10594 Visrag: Vision-based retrieval-augmented generation on multi-modality documents . Preprint, arXiv:2410.10594

    2024 arXiv

  25. [33]

    Yuan Zhang, Chun-Kai Fan, Junpeng Ma, Wenzhao Zheng, Tao Huang, Kuan Cheng, Denis Gudovskiy, Tomoyuki Okuno, Yohei Nakata, Kurt Keutzer, and Shanghang Zhang. 2025. https://arxiv.org/abs/2410.04417 Sparsevlm: Visual token sparsification for efficient vision-language model infer...

    2025 arXiv

  26. [34]

    Fengbin Zhu, Wenqiang Lei, Fuli Feng, Chao Wang, Haozhou Zhang, and Tat-Seng Chua. 2022. Towards complex document understanding by discrete reasoning. In Proceedings of the 30th ACM International Conference on Multimedia, pages 4857--4866

    2022

  27. [35]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list...

  28. [36]

    write newline

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

Pith tools

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