Pith. sign in

REVIEW 4 major objections 3 minor 54 references

A vision-language model using block discrete diffusion matches autoregressive OCR accuracy while decoding in parallel, achieving roughly three times the throughput.

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 · deepseek-v4-flash

2026-08-02 22:23 UTC pith:Z62RGGWL

load-bearing objection Block diffusion for OCR works in practice and the failure of global masked diffusion on dense text is a real finding, but the paper overclaims the comparison to autoregressive models and has a few sloppy numbers that need fixing. the 4 major comments →

arxiv 2602.16872 v2 pith:Z62RGGWL submitted 2026-02-18 cs.CV

DODO: Discrete OCR Diffusion Models

Sean Man , Gilad Deutch , Roy Ganz , Roi Ronen , Shahar Tsiper , Shai Mazor , Niv Nayman This is my paper
classification cs.CV
keywords OCRdiscrete diffusionmasked diffusionblock diffusionvision-language modelparallel decodingdocument transcriptionKV-caching
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.

The paper argues that OCR's determinism is the key to parallel decoding: because an image dictates a single correct transcription, masked diffusion's assumption that tokens can be predicted independently actually holds, unlike in captioning. It then shows why vanilla masked diffusion still fails at OCR: the rigid, exact-match output cannot tolerate length mismatches or positional drift that flexible text generation absorbs. To fix this, DODO decomposes generation into causally anchored blocks, so each block denoises in parallel against a committed prefix. The result is near-autoregressive accuracy on document benchmarks, with a block-causal variant using KV-caching achieving about 3x the throughput of the autoregressive backbone.

Core claim

DODO is the first vision-language model to apply block discrete diffusion to OCR. The central discovery is that block-level autoregressive anchoring resolves the structural instability of global masked diffusion: length estimation errors and absolute positional misalignment, which are catastrophic for exact transcription, are bounded within each block, while tokens inside a block are still decoded in parallel. Using 256-token blocks and training with block-causal masking, DODO achieves normalized edit distances of 0.066 on OmniDocBench and 0.041 on Fox-Page-EN, surpassing its autoregressive backbone, and DODO-fast reaches about 63 tokens per second, roughly 3x the backbone's throughput.

What carries the argument

Block discrete diffusion: the sequence is split into contiguous blocks of up to 256 tokens; each block is decoded by masked diffusion in parallel, conditioned on the already-decoded prefix blocks, replicating autoregressive causality at block granularity. The block-causal attention variant freezes prefix representations and enables exact KV-caching; the bidirectional variant recomputes representations for maximal information flow. Ablations show that this structure, not the OCR data alone, is responsible for the gains — full-sequence masked diffusion fails even with oracle length guidance.

Load-bearing premise

The empirical comparison assumes that the off-the-shelf autoregressive baselines are fair stand-ins for DODO, even though they were not fine-tuned on the same OCR corpus; without that assumption, the accuracy and throughput claims may partly reflect data rather than the block-diffusion design.

What would settle it

Fine-tune the same autoregressive backbone on the same OCR training corpus and measure normalized edit distance and tokens per second on the same benchmarks; if the autoregressive model matches DODO's accuracy, the paper's central accuracy claim collapses. A second check: run DODO on an OCR set with genuinely ambiguous transcriptions (e.g., handwriting); if errors rise sharply where conditional independence fails, the mechanism is confirmed.

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

If this is right

  • OCR no longer requires one forward pass per token; latency decouples from sequence length.
  • Block-causal training makes exact KV-caching compatible with diffusion, unlocking roughly 3x throughput.
  • Diffusion VLMs can match autoregressive accuracy on dense document transcription, closing a gap that prior diffusion VLMs left open.
  • Larger block sizes than used in text-only block diffusion are viable because OCR's low-entropy posterior supports conditional independence.
  • The structural benefit comes from training: applying block decoding at inference time to a vanilla masked diffusion model does not recover performance.

Where Pith is reading between the lines

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

  • If the result generalizes, any deterministic serialization task with exact-match targets (structured data extraction, code from screenshots, formula transcription) is a candidate for block diffusion with similar speedups.
  • The fairness caveat: the reported accuracy superiority over the autoregressive backbone may partly reflect training data rather than architecture; an autoregressive model fine-tuned identically is the natural check.
  • The accuracy/throughput tradeoff with block size suggests that adaptive block sizing or recurrent refinement of low-confidence tokens could push both metrics; the paper leaves this implicit.
  • DODO-fast's inverse block-size trend (smaller blocks better with frozen context) hints at a testable hypothesis about representation drift that could guide future cache-friendly architectures.

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

4 major / 3 minor

Summary. The paper introduces DODO, a 3B VLM that applies block discrete diffusion to OCR. DODO trains on olmOCR-mix-1025 and is evaluated on OmniDocBench and Fox-Page-EN. It comes in two variants: DODO (full bidirectional attention) and DODO-fast (block-causal attention with exact KV-caching). The paper claims that DODO achieves near-SOTA accuracy while enabling up to 3x faster inference than an autoregressive Qwen2.5-VL backbone, and argues that block-wise generation is necessary to avoid the alignment failures of global masked diffusion on the rigid, deterministic OCR task.

Significance. If the central claims hold, the contribution is valuable: it is the first demonstration that block discrete diffusion can be made to work in a multimodal OCR setting, and the ablation showing vanilla masked diffusion collapses while block training succeeds provides a concrete, falsifiable structural story. The DODO-fast variant also shows a practical speed/accuracy trade-off via exact KV-caching. However, the comparisons against autoregressive baselines are confounded by training data, and the throughput/accuracy framing needs sharper reporting. The paper's methodological core is sound enough to warrant revision rather than rejection.

major comments (4)
  1. [§5.2, Table 1] The headline comparison against Qwen2.5-VL is confounded: DODO is fine-tuned for 200k steps on olmOCR-mix-1025 (Section A), while the Qwen2.5-VL baselines are off-the-shelf. The claim that DODO 'outperforms its autoregressive backbone' cannot separate architecture from data. Table 2 only ablates MDM variants; it does not include an AR model fine-tuned on the same data and training budget. A matched AR baseline is load-bearing for the §5.3 conclusion that discrete diffusion is a viable high-performance alternative to AR decoding.
  2. [§5.4, Figure 5; Abstract] The speedup claim is internally inconsistent. Figure 5 reports DODO at 22.90 TPS vs Qwen2.5-VL-3B at 21.00 TPS (≈1.09x), while DODO-fast reaches 63.18 TPS (≈3.0x). The abstract in the front matter states 'up to 5x faster inference', while the body abstract states '3×'; no configuration supports 5x. Moreover, the 3x speedup belongs to DODO-fast, whose NED is 0.159 vs 0.066 for DODO — a 2.4x error increase. The paper must clearly separate 'near-SOTA accuracy' from '3x speedup' and correct the abstract.
  3. [§6.1, Table 2] The text claims DODO reduces error rates by 'nearly 10× compared to the best Vanilla configuration'. The best Vanilla row is 'Vanilla Oracle' at 0.100 NED; DODO at 0.066 is only a ~1.5x improvement. The ~10x figure only holds against the non-oracle Vanilla 8192 row (0.834). This misstatement weakens the structural-advantage narrative and must be corrected with the appropriate baseline clearly specified.
  4. [§5.4, §6.2] Throughput numbers are reported as single point values without measurement conditions (batch size, input length, hardware, number of runs) or error bars. Given that the central claim is an efficiency advantage and the differences between configurations are sometimes small (e.g., 22.9 vs 22.0 TPS in Table 3), the authors should specify the protocol and, where feasible, report variance across runs.
minor comments (3)
  1. [Abstract] The two abstract versions disagree on the speedup factor ('5x' in the header, '3×' in the body). Please unify after correcting the actual supported number.
  2. [§5.1, Table 2] The 'Vanilla Oracle' row is under-specified: the table lists Max Length and Inference Block Size as '-', and the caption does not explain how the oracle length is provided. Add a footnote.
  3. [Throughout] Variant names are inconsistently typeset ('DODOfast', 'DODO fast', 'DODOfast'). Use a single notation. Minor typos such as 'suffers' in §5.4 and 'suffers a sharp accuracy collapse' in §6.2 can be cleaned up.

Circularity Check

0 steps flagged

No significant circularity; the core block-diffusion claim rests on a same-data ablation, while unmatched AR baselines are a confound rather than a definitional reduction.

full rationale

The derivation chain is not circular. DODO's mechanism (block discrete diffusion with block-causal masking) is imported from external prior work (Arriola et al. 2025; Wu et al. 2025a,b) and adapted to VLMs, so the core architectural claim does not reduce to a self-citation. The central structural claim—that block training, not data, explains accuracy—is supported by Table 2, which trains Vanilla MDM and Block MDM on the same olmOCR-mix-1025 corpus and compares NED; that is a controlled internal ablation rather than a fitted prediction. Throughput figures in Fig. 5 are measured, not derived from fitted parameters. No parameter is fit to OmniDocBench/Fox-Page and then 'predicted'; no uniqueness theorem is invoked; no ansatz is smuggled via self-citation. The main weakness is experimental: Table 1 compares DODO (fine-tuned on olmOCR-mix-1025, 200k steps) against off-the-shelf Qwen2.5-VL AR models, so the 'outperforms its autoregressive backbone' claim conflates architecture with OCR fine-tuning. This is a validity/confound issue, not circularity. Likewise, the text's '~10x reduction' versus Table 2's best Vanilla Oracle 0.100 appears numerically inconsistent (0.066/0.100 ≈ 1.5x), but an arithmetic inconsistency is not a definitional reduction. Background self-citations (Ganz et al. 2023/2024, Abramovich et al. 2024, Ronen et al. 2022, Nacson et al. 2025) appear only in the introduction and related work and are not load-bearing for the diffusion derivation.

Axiom & Free-Parameter Ledger

3 free parameters · 5 axioms · 0 invented entities

The central claim rests on a domain assumption about OCR being deterministic and conditionally independent, plus several empirically tuned structural hyperparameters. No new theoretical entities are postulated. The most load-bearing assumption is the fairness of comparing against non-fine-tuned AR baselines.

free parameters (3)
  • Confidence threshold p = 0.98
    Default unmasking threshold chosen after the Section B.1 ablation; it controls the accuracy/speed trade-off and is not fixed by theory.
  • Block size L' = 256 for DODO; 32 for the best reported DODO-fast
    Ablated in Table 3; accuracy is non-monotonic in block size. The main text stresses 256, but Table 1's DODO-fast NED matches the block-32 row of Table 3.
  • Maximum sequence length = 8192
    Canvas length for dense documents; affects the length-mismatch failure and compute cost, chosen without a principled rule.
axioms (5)
  • domain assumption OCR posterior is effectively unimodal, so token predictions are approximately conditionally independent given the image (Eq. 4).
    Argued in Section 4.1 rather than derived; real documents contain ambiguity in punctuation, layout, and hyphenation.
  • domain assumption The block factorization p(x1:L|I,c) = prod_b p(x^(b)|x^(<b),I,c) (Eq. 5) with block size up to 256 preserves accuracy.
    Empirically supported by Table 3, but only on two English benchmarks and with degradation at 512/1024 and for the static-prefix fast variant.
  • standard math Block-causal attention makes exact KV-caching valid.
    Follows from the mask preventing prefix tokens from attending to future blocks; Section 5.1 and Figure A1.
  • standard math Standard masked-diffusion training with carry-over unmasking is the correct learning objective.
    Adopted from Sahoo et al. 2024a and prior work; the paper does not revisit the no-revision constraint.
  • domain assumption NED on OmniDocBench-English and Fox-Page-EN is a sufficient proxy for OCR quality.
    Only two English benchmarks are used; there is no multilingual or handwritten-text evaluation, and no significance testing.

pith-pipeline@v1.3.0-alltime-deepseek · 34225 in / 11535 out tokens · 104665 ms · 2026-08-02T22:23:59.911881+00:00 · methodology

0 comments
read the original abstract

Optical Character Recognition (OCR) is a fundamental task for digitizing information, serving as a critical bridge between visual data and textual understanding. While modern Vision-Language Models (VLM) have achieved high accuracy in this domain, they predominantly rely on autoregressive decoding, which becomes computationally expensive and slow for long documents as it requires a sequential forward pass for every generated token. We identify a key opportunity to overcome this bottleneck: unlike open-ended generation, OCR is a highly deterministic task where the visual input strictly dictates a unique output sequence, theoretically enabling efficient, parallel decoding via diffusion models. However, we show that existing masked diffusion models fail to harness this potential; those introduce structural instabilities that are benign in flexible tasks, like captioning, but catastrophic for the rigid, exact-match requirements of OCR. To bridge this gap, we introduce DODO, the first VLM to utilize block discrete diffusion and unlock its speedup potential for OCR. By decomposing generation into blocks, DODO mitigates the synchronization errors of global diffusion. Empirically, our method achieves near state-of-the-art accuracy while enabling up to 5x faster inference compared to autoregressive baselines.

Figures

Figures reproduced from arXiv: 2602.16872 by Gilad Deutch, Niv Nayman, Roi Ronen, Roy Ganz, Sean Man, Shahar Tsiper, Shai Mazor.

Figure 1
Figure 1. Figure 1: DODO: High-throughput parallel generation. Unlike autoregressive models constrained to a strict left-to-right sequence, DODO generates text across the entire canvas simultaneously (with same color) based on visual confidence. In this example, it resolves 148 tokens in just 15 forward passes (≈ 10 tokens/step on average). Notably, large, distinct regions appear early, while ambiguous high￾frequency tokens (… view at source ↗
Figure 2
Figure 2. Figure 2: Semantically flexible vs. semantically rigid vision–language tasks. Left: Image captioning admits multiple, semantically equivalent descriptions of the same image. Different decoding trajectories can converge to distinct but equally valid captions, and lexical or structural variations are naturally absorbed. Right: OCR requires a single, exact transcription determined by the image. Even minimal local devia… view at source ↗
Figure 3
Figure 3. Figure 3: Conditional independence assumption. Parallel de￾coding assumes masked that masked tokens can be predicted inde￾pendently given the context. (Top) In open-ended tasks, ambiguity between valid options (e.g., “Eiffel Tower” vs. “Great Wall”) risks sampling incoherent mixtures like “Eiffel Wall.” (Bottom) In deter￾ministic regimes like OCR, the strong visual signal resolves this ambiguity, enabling conflict-f… view at source ↗
Figure 4
Figure 4. Figure 4: Full vs. block diffusion. In standard full diffusion (left), MDM sampling is applied globally to the entire sequence. In contrast, block diffusion (right) restricts parallel sampling to discrete windows, processing blocks sequentially from left to right. segments collide – a fundamental challenge that limits the efficacy of purely parallel OCR. 4.3. Block Diffusion as a Structural Remedy Block discrete dif… view at source ↗
Figure 5
Figure 5. Figure 5: Inference throughput comparison. While standard DODO matches the speed of the autoregressive Qwen 2.5 VL baseline (≈ 21 tokens/sec), the DODO fast leverages block-causal attention and KV-caching to triple the throughput to ≈ 63 token￾s/sec, establishing a new efficiency standard for diffusion-based VLMs. DODO benefits from specialized OCR training, we demon￾strate in our ablations (Section 6.1) that data d… view at source ↗
Figure 6
Figure 6. Figure 6: Decoding Efficiency. Distribution of inference steps normalized by output length. The autoregressive baseline is struc￾turally limited to generating a single token per step. In contrast, DODO leverages parallel decoding to generate multiple tokens simultaneously, effectively compressing the inference process by an order of magnitude (typically < 0.1 steps per token). more tokens in parallel reduces the tot… 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

54 extracted references · 26 linked inside Pith

  1. [1]

    write newline

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

  2. [2]

    https://mistral.ai/news/mistral-ocr, 2025

    Mistral ocr. https://mistral.ai/news/mistral-ocr, 2025. Mistral AI Optical Character Recognition model

    2025

  3. [3]

    Clipter: Looking at the bigger picture in scene text recognition

    Aberdam, A., Bensa \" d, D., Golts, A., Ganz, R., Nuriel, O., Tichauer, R., Mazor, S., and Litman, R. Clipter: Looking at the bigger picture in scene text recognition. In Proceedings of the IEEE/CVF international conference on computer vision, pp.\ 21706--21717, 2023

    2023

  4. [4]

    Visfocus: Prompt-guided vision encoders for ocr-free dense document understanding

    Abramovich, O., Nayman, N., Fogel, S., Lavi, I., Litman, R., Tsiper, S., Tichauer, R., Appalaraju, S., Mazor, S., and Manmatha, R. Visfocus: Prompt-guided vision encoders for ocr-free dense document understanding. In European Conference on Computer Vision, pp.\ 241--259. Springer, 2024

    2024

  5. [5]

    L., Borgeaud, S., Brock, A., Nematzadeh, A., Sharifzadeh, S., Bi\' n kowski, M

    Alayrac, J.-B., Donahue, J., Luc, P., Miech, A., Barr, I., Hasson, Y., Lenc, K., Mensch, A., Millican, K., Reynolds, M., Ring, R., Rutherford, E., Cabi, S., Han, T., Gong, Z., Samangooei, S., Monteiro, M., Menick, J. L., Borgeaud, S., Brock, A., Nematzadeh, A., Sharifzadeh, S., Bi\' n kowski, M. a., Barreira, R., Vinyals, O., Zisserman, A., and Simonyan, ...

    2022

  6. [6]

    L., and Parikh, D

    Antol, S., Agrawal, A., Lu, J., Mitchell, M., Batra, D., Zitnick, C. L., and Parikh, D. Vqa: Visual question answering. In Proceedings of the IEEE international conference on computer vision, pp.\ 2425--2433, 2015

    2015

  7. [7]

    T., Yang, Z., Qi, Z., Han, J., Sahoo, S

    Arriola, M., Gokaslan, A., Chiu, J. T., Yang, Z., Qi, Z., Han, J., Sahoo, S. S., and Kuleshov, V. Block diffusion: Interpolating between autoregressive and diffusion language models. arXiv preprint arXiv:2503.09573, 2025

    Pith/arXiv arXiv 2025

  8. [8]

    D., Ho, J., Tarlow, D., and van den Berg, R

    Austin, J., Johnson, D. D., Ho, J., Tarlow, D., and van den Berg, R. Structured denoising diffusion models in discrete state-spaces. Advances in Neural Information Processing Systems, 2021

    2021

  9. [9]

    Parallel sampling from masked diffusion models via conditional independence testing, 2025

    Azangulov, I., Pandeva, T., Prasad, N., Zazo, J., and Karmalkar, S. Parallel sampling from masked diffusion models via conditional independence testing, 2025. URL https://arxiv.org/abs/2510.21961

    arXiv 2025

  10. [10]

    Qwen3-vl technical report, 2025 a

    Bai, S., Cai, Y., Chen, R., Chen, K., Chen, X., Cheng, Z., Deng, L., Ding, W., Gao, C., Ge, C., Ge, W., Guo, Z., Huang, Q., Huang, J., Huang, F., Hui, B., Jiang, S., Li, Z., Li, M., Li, M., Li, K., Lin, Z., Lin, J., Liu, X., Liu, J., Liu, C., Liu, Y., Liu, D., Liu, S., Lu, D., Luo, R., Lv, C., Men, R., Meng, L., Ren, X., Ren, X., Song, S., Sun, Y., Tang, ...

    Pith/arXiv arXiv 2025

  11. [12]

    Bai, S., Chen, K., Liu, X., Wang, J., Ge, W., Song, S., Dang, K., Wang, P., Wang, S., Tang, J., et al. Qwen2. 5-vl technical report. arXiv preprint arXiv:2502.13923, 2025 c

    Pith/arXiv arXiv 2025

  12. [13]

    Nougat: Neural optical understanding for academic documents, 2023

    Blecher, L., Cucurull, G., Scialom, T., and Stojnic, R. Nougat: Neural optical understanding for academic documents, 2023. URL https://arxiv.org/abs/2308.13418

    Pith/arXiv arXiv 2023

  13. [14]

    Sharegpt4v: Improving large multi-modal models with better captions, 2023

    Chen, L., Li, J., Dong, X., Zhang, P., He, C., Wang, J., Zhao, F., and Lin, D. Sharegpt4v: Improving large multi-modal models with better captions, 2023. URL https://arxiv.org/abs/2311.12793

    Pith/arXiv arXiv 2023

  14. [15]

    Chen, M., Tworek, J., Jun, H., Yuan, Q., de Oliveira Pinto, H. P., Kaplan, J., Edwards, H., Burda, Y., Joseph, N., Brockman, G., Ray, A., Puri, R., Krueger, G., Petrov, M., Khlaaf, H., Sastry, G., Mishkin, P., Chan, B., Gray, S., Ryder, N., Pavlov, M., Power, A., Kaiser, L., Bavarian, M., Winter, C., Tillet, P., Such, F. P., Cummings, D., Plappert, M., Ch...

    2021

  15. [16]

    Expanding performance boundaries of open-source multimodal models with model, data, and test-time scaling, 2025

    Chen, Z., Wang, W., Cao, Y., Liu, Y., Gao, Z., Cui, E., Zhu, J., Ye, S., Tian, H., Liu, Z., Gu, L., Wang, X., Li, Q., Ren, Y., Chen, Z., Luo, J., Wang, J., Jiang, T., Wang, B., He, C., Shi, B., Zhang, X., Lv, H., Wang, Y., Shao, W., Chu, P., Tu, Z., He, T., Wu, Z., Deng, H., Ge, J., Chen, K., Zhang, K., Wang, L., Dou, M., Lu, L., Zhu, X., Lu, T., Lin, D.,...

    Pith/arXiv arXiv 2025

  16. [17]

    Training verifiers to solve math word problems

    Cobbe, K., Kosaraju, V., Bavarian, M., Chen, M., Jun, H., Kaiser, L., Plappert, M., Tworek, J., Hilton, J., Nakano, R., et al. Training verifiers to solve math word problems. arXiv preprint arXiv:2110.14168, 2021

    Pith/arXiv arXiv 2021

  17. [18]

    N., and Hoi, S

    Dai, W., Li, J., LI, D., Tiong, A., Zhao, J., Wang, W., Li, B., Fung, P. N., and Hoi, S. Instructblip: Towards general-purpose vision-language models with instruction tuning. In Oh, A., Naumann, T., Globerson, A., Saenko, K., Hardt, M., and Levine, S. (eds.), Advances in Neural Information Processing Systems, volume 36, pp.\ 49250--49267. Curran Associate...

    2023

  18. [19]

    Towards models that can see and read

    Ganz, R., Nuriel, O., Aberdam, A., Kittenplon, Y., Mazor, S., and Litman, R. Towards models that can see and read. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pp.\ 21718--21728, October 2023

    2023

  19. [20]

    B., Nuriel, O., Mazor, S., and Litman, R

    Ganz, R., Kittenplon, Y., Aberdam, A., Avraham, E. B., Nuriel, O., Mazor, S., and Litman, R. Question aware vision transformer for multimodal reasoning, 2024. URL https://arxiv.org/abs/2402.05472

    Pith/arXiv arXiv 2024

  20. [21]

    Llava-onevision: Easy visual task transfer, 2024

    Li, B., Zhang, Y., Guo, D., Zhang, R., Li, F., Zhang, H., Zhang, K., Zhang, P., Li, Y., Liu, Z., and Li, C. Llava-onevision: Easy visual task transfer, 2024. URL https://arxiv.org/abs/2408.03326

    Pith/arXiv arXiv 2024

  21. [22]

    BLIP : Bootstrapping language-image pre-training for unified vision-language understanding and generation

    Li, J., Li, D., Xiong, C., and Hoi, S. BLIP : Bootstrapping language-image pre-training for unified vision-language understanding and generation. In Chaudhuri, K., Jegelka, S., Song, L., Szepesvari, C., Niu, G., and Sabato, S. (eds.), Proceedings of the 39th International Conference on Machine Learning, volume 162 of Proceedings of Machine Learning Resear...

    2022

  22. [23]

    Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models, 2023

    Li, J., Li, D., Savarese, S., and Hoi, S. Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models, 2023. URL https://arxiv.org/abs/2301.12597

    Pith/arXiv arXiv 2023

  23. [24]

    Lavida: A large diffusion model for vision-language understanding

    Li, S., Kallidromitis, K., Bansal, H., Gokul, A., Kato, Y., Kozuka, K., Kuen, J., Lin, Z., Chang, K.-W., and Grover, A. Lavida: A large diffusion model for vision-language understanding. In The Thirty-ninth Annual Conference on Neural Information Processing Systems, 2025 a

    2025

  24. [25]

    dots.ocr: Multilingual document layout parsing in a single vision-language model, 2025 b

    Li, Y., Yang, G., Liu, H., Wang, B., and Zhang, C. dots.ocr: Multilingual document layout parsing in a single vision-language model, 2025 b . URL https://arxiv.org/abs/2512.02498

    arXiv 2025

  25. [26]

    Monkeyocr: A unified ocr system with multi-stage pipelines

    Li, Z., Liu, Y., Liu, Q., Ma, Z., Zhang, Z., Zhang, S., Guo, Z., Zhang, J., Wang, X., and Bai, X. Monkeyocr: A unified ocr system with multi-stage pipelines. arXiv preprint arXiv:2506.05218, 2025 c

    arXiv 2025

  26. [27]

    Lin, T.-Y., Maire, M., Belongie, S., Hays, J., Perona, P., Ramanan, D., Doll \'a r, P., and Zitnick, C. L. Microsoft coco: Common objects in context. In European conference on computer vision, pp.\ 740--755. Springer, 2014

    2014

  27. [28]

    Liu, C. et al. Focus anywhere for fine-grained multi-page document understanding. arXiv preprint arXiv:2405.14295, 2024

    Pith/arXiv arXiv 2024

  28. [29]

    Nvila: Efficient frontier visual language models, 2025

    Liu, Z., Zhu, L., Shi, B., Zhang, Z., Lou, Y., Yang, S., Xi, H., Cao, S., Gu, Y., Li, D., Li, X., Fang, Y., Chen, Y., Hsieh, C.-Y., Huang, D.-A., Cheng, A.-C., Nath, V., Hu, J., Liu, S., Krishna, R., Xu, D., Wang, X., Molchanov, P., Kautz, J., Yin, H., Han, S., and Lu, Y. Nvila: Efficient frontier visual language models, 2025. URL https://arxiv.org/abs/2412.04468

    Pith/arXiv arXiv 2025

  29. [30]

    Plan for speed--dilated scheduling for masked diffusion language models

    Luxembourg, O., Permuter, H., and Nachmani, E. Plan for speed--dilated scheduling for masked diffusion language models. arXiv preprint arXiv:2506.19037, 2025

    Pith/arXiv arXiv 2025

  30. [31]

    Nanonets-ocr-s: A model for transforming documents into structured markdown with intelligent content recognition and semantic tagging, 2025

    Mandal, S., Talewar, A., Ahuja, P., and Juvatkar, P. Nanonets-ocr-s: A model for transforming documents into structured markdown with intelligent content recognition and semantic tagging, 2025

    2025

  31. [32]

    Docvqa: A dataset for vqa on document images

    Mathew, M., Karatzas, D., and Jawahar, C. Docvqa: A dataset for vqa on document images. In Proceedings of the IEEE/CVF winter conference on applications of computer vision, pp.\ 2200--2209, 2021

    2021

  32. [33]

    S., Aberdam, A., Ganz, R., Ben Avraham, E., Golts, A., Kittenplon, Y., Mazor, S., and Litman, R

    Nacson, M. S., Aberdam, A., Ganz, R., Ben Avraham, E., Golts, A., Kittenplon, Y., Mazor, S., and Litman, R. Docvlm: Make your vlm an efficient reader. In Proceedings of the Computer Vision and Pattern Recognition Conference, pp.\ 29005--29015, 2025

    2025

  33. [34]

    T., Kim, Y., Gurbuz, A

    Nassar, A., Marafioti, A., Omenetti, M., Lysak, M., Livathinos, N., Auer, C., Morin, L., de Lima, R. T., Kim, Y., Gurbuz, A. S., Dolfi, M., Farré, M., and Staar, P. W. J. Smoldocling: An ultra-compact vision-language model for end-to-end multi-modal document conversion, 2025. URL https://arxiv.org/abs/2503.11576

    Pith/arXiv arXiv 2025

  34. [35]

    Ouyang, L. et al. Omnidocbench: Benchmarking diverse pdf document parsing with comprehensive annotations. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025

    2025

  35. [36]

    Poznanski, A. et al. olmocr: Unlocking trillions of tokens in pdfs with vision language models. Hugging Face Datasets, 2024. https://huggingface.co/datasets/allenai/olmOCR-mix-1025

    2024

  36. [37]

    olmocr: Unlocking trillions of tokens in pdfs with vision language models

    Poznanski, J., Rangapur, A., Borchardt, J., Dunkelberger, J., Huff, R., Lin, D., Wilhelm, C., Lo, K., and Soldaini, L. olmocr: Unlocking trillions of tokens in pdfs with vision language models. arXiv preprint arXiv:2502.18443, 2025

    arXiv 2025

  37. [38]

    Glass: Global to local attention for scene-text spotting

    Ronen, R., Tsiper, S., Anschel, O., Lavi, I., Markovitz, A., and Manmatha, R. Glass: Global to local attention for scene-text spotting. In European Conference on Computer Vision, pp.\ 249--266. Springer, 2022

    2022

  38. [39]

    Simple and effective masked diffusion language models

    Sahoo, S., Arriola, M., Schiff, Y., Gokaslan, A., Marroquin, E., Chiu, J., Rush, A., and Kuleshov, V. Simple and effective masked diffusion language models. Advances in Neural Information Processing Systems, 37: 0 130136--130184, 2024 a

    2024

  39. [40]

    S., Arriola, M., Schiff, Y., Gokaslan, A., Marroquin, E., Chiu, J

    Sahoo, S. S., Arriola, M., Schiff, Y., Gokaslan, A., Marroquin, E., Chiu, J. T., Rush, A., and Kuleshov, V. Simple and effective masked diffusion language models, 2024 b . URL https://arxiv.org/abs/2406.07524

    Pith/arXiv arXiv 2024

  40. [41]

    Simplified and Generalized Masked Diffusion for Discrete Data

    Shi, J., Han, K., Wang, Z., Doucet, A., and Titsias, M. Simplified and Generalized Masked Diffusion for Discrete Data . Advances in Neural Information Processing Systems, 37: 0 103131--103167, December 2024

    2024

  41. [42]

    Textcaps: a dataset for image captioning with reading comprehension, 2020

    Sidorov, O., Hu, R., Rohrbach, M., and Singh, A. Textcaps: a dataset for image captioning with reading comprehension, 2020. URL https://arxiv.org/abs/2003.12462

    Pith/arXiv arXiv 2020

  42. [43]

    Towards vqa models that can read, 2019

    Singh, A., Natarajan, V., Shah, M., Jiang, Y., Chen, X., Batra, D., Parikh, D., and Rohrbach, M. Towards vqa models that can read, 2019. URL https://arxiv.org/abs/1904.08920

    Pith/arXiv arXiv 2019

  43. [44]

    Mineru: An open-source solution for precise document content extraction

    Wang, B., Xu, C., Zhao, X., Ouyang, L., Wu, F., Zhao, Z., Xu, R., Liu, K., Qu, Y., Shang, F., et al. Mineru: An open-source solution for precise document content extraction. arXiv preprint arXiv:2409.18839, 2024

    Pith/arXiv arXiv 2024

  44. [45]

    From object detection to text detection and recognition: A brief evolution history of optical character recognition

    Wang, H., Pan, C., Guo, X., Ji, C., and Deng, K. From object detection to text detection and recognition: A brief evolution history of optical character recognition. Wiley Interdisciplinary Reviews: Computational Statistics, 13 0 (5): 0 e1547, 2021

    2021

  45. [46]

    General ocr theory: Towards ocr-2.0 via a unified end-to-end model, 2024

    Wei, H., Liu, C., Chen, J., Wang, J., Kong, L., Xu, Y., Ge, Z., Zhao, L., Sun, J., Peng, Y., Han, C., and Zhang, X. General ocr theory: Towards ocr-2.0 via a unified end-to-end model, 2024. URL https://arxiv.org/abs/2409.01704

    Pith/arXiv arXiv 2024

  46. [48]

    Deepseek-ocr: Contexts optical compression, 2025 b

    Wei, H., Sun, Y., and Li, Y. Deepseek-ocr: Contexts optical compression, 2025 b . URL https://arxiv.org/abs/2510.18234

    Pith/arXiv arXiv 2025

  47. [49]

    Fast-dllm v2: Efficient block-diffusion llm

    Wu, C., Zhang, H., Xue, S., Diao, S., Fu, Y., Liu, Z., Molchanov, P., Luo, P., Han, S., and Xie, E. Fast-dllm v2: Efficient block-diffusion llm. arXiv preprint arXiv:2509.26328, 2025 a

    arXiv 2025

  48. [50]

    Fast-dllm: Training-free acceleration of diffusion language models

    Wu, C., Zhang, H., Xue, S., Liu, Z., Diao, S., Zhu, L., Luo, P., Han, S., and Xie, E. Fast-dllm: Training-free acceleration of diffusion language models. arXiv preprint arXiv:2505.22618, 2025 b

    Pith/arXiv arXiv 2025

  49. [51]

    Deepseek-vl2: Mixture-of-experts vision-language models for advanced multimodal understanding, 2024

    Wu, Z., Chen, X., Pan, Z., Liu, X., Liu, W., Dai, D., Gao, H., Ma, Y., Wu, C., Wang, B., Xie, Z., Wu, Y., Hu, K., Wang, J., Sun, Y., Li, Y., Piao, Y., Guan, K., Liu, A., Xie, X., You, Y., Dong, K., Yu, X., Zhang, H., Zhao, L., Wang, Y., and Ruan, C. Deepseek-vl2: Mixture-of-experts vision-language models for advanced multimodal understanding, 2024. URL ht...

    Pith/arXiv arXiv 2024

  50. [52]

    Llada-v: Large language diffusion models with visual instruction tuning

    You, Z., Nie, S., Zhang, X., Hu, J., Zhou, J., Lu, Z., Wen, J.-R., and Li, C. Llada-v: Large language diffusion models with visual instruction tuning. arXiv preprint arXiv:2505.16933, 2025

    Pith/arXiv arXiv 2025

  51. [53]

    Dimple: Discrete diffusion multimodal large language model

    Yu, R., Ma, X., and Wang, X. Dimple: Discrete diffusion multimodal large language model. arXiv preprint arXiv:2505.16990, 2025

    Pith/arXiv arXiv 2025

  52. [54]

    Mmmu: A massive multi-discipline multimodal understanding and reasoning benchmark for expert agi, 2024

    Yue, X., Ni, Y., Zhang, K., Zheng, T., Liu, R., Zhang, G., Stevens, S., Jiang, D., Ren, W., Sun, Y., Wei, C., Yu, B., Yuan, R., Sun, R., Yin, M., Zheng, B., Yang, Z., Liu, Y., Huang, W., Sun, H., Su, Y., and Chen, W. Mmmu: A massive multi-discipline multimodal understanding and reasoning benchmark for expert agi, 2024. URL https://arxiv.org/abs/2311.16502

    Pith/arXiv arXiv 2024

  53. [55]

    A reparameterized discrete diffusion model for text generation

    Zheng, L., Yuan, J., Yu, L., and Kong, L. A reparameterized discrete diffusion model for text generation. arXiv preprint arXiv:2302.05737, 2023

    Pith/arXiv arXiv 2023

  54. [56]

    Dream: Diffusion rectification and estimation-adaptive models

    Zhou, J., Ding, T., Chen, T., Jiang, J., Zharkov, I., Zhu, Z., and Liang, L. Dream: Diffusion rectification and estimation-adaptive models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 8342--8351, 2024

    2024