Pith. sign in

REVIEW 4 major objections 5 minor 119 references

A single training-free sampler can generate text and image concurrently, letting each modality correct the other in every denoising step, and it beats uncoupled parallel decoders on joint editing and visual-reasoning benchmarks.

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 05:54 UTC pith:7DER3Q5I

load-bearing objection Useful empirical work and new datasets, but the within-step coupling that is the paper's headline mechanism is not actually in the algorithm: the image branch conditions on the previous text state, not the state just sampled. the 4 major comments →

arxiv 2607.13188 v1 pith:7DER3Q5I submitted 2026-07-14 cs.LG

Concurrent Image Understanding and Generation: Self-Correcting Coupled Markov Jump Processes

Minh-Quan Le , Armand Comas , Alexandros Lattas , Stylianos Moschoglou , Pedro V\'elez , Amit Raj , Aaron Germuth , Thabo Beeler
show 2 more authors
Dimitris Samaras Di Qiu
This is my paper
classification cs.LG
keywords masked diffusion modelsjoint multimodal generationMarkov jump processesremaskingself-correctioncross-modal attentionimage editingvisual reasoning
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.

Text and image generation are usually run sequentially or in parallel branches that only share history. This paper claims that a frozen masked diffusion model can instead generate both modalities concurrently, in a single forward pass per step, by coupling each modality's unmasking decisions to the other's confidence through cross-modal attention, and by remasking tokens the other modality contradicts. The resulting sampler, CO2Jump, is claimed to achieve the best joint performance on image editing plus understanding and on two visual-reasoning tasks (maze and nonogram solving), and to improve monotonically as more denoising steps are used. If true, concurrent generation with self-correction is available without retraining or an auxiliary evaluator, and the benefits of coupling compound across the trajectory.

Core claim

The paper's central claim is that cross-modal coupling can be extracted from a frozen jointly-trained masked diffusion model at inference time, without extra forward passes. It rewrites the joint reverse step using the chain rule: text updates first from the joint state, and the image update should condition on the freshly sampled text. Since that fresh text is unavailable without a second forward pass, the image side instead reads the model's own text self-confidence, propagates it through cross-modal attention, and mixes it with image self-confidence under an entropy-derived gate. A death jump then remasks the lowest-confidence committed tokens and a birth jump reveals the highest-confiden

What carries the argument

The load-bearing object is the Coupled Confidence score. After one forward pass, every position gets a Self-Confidence from Gumbel-max sampling of the model's belief. Text positions are scored by self-confidence alone, because the text update needs no cross-modal information, whereas image positions are scored by an entropy-gated blend of self-confidence and a CrossSignal: the text self-confidences weighted by image-to-text cross-attention from the same hidden states, with a mask-aware bias. An entropy gate λ = H_image/(H_image+H_text) decides how much the image defers to text. Scores are rank-normalized within each modality so vocabularies are comparable. These scores drive a death-birth ju

Load-bearing premise

The image branch's cross-modal correction is actually a surrogate — attention-weighted text beliefs from the current forward pass stand in for the text tokens that will only be sampled in this step; if that proxy does not track the real next-step text state, the coupling gives no true conditioning and the reported gains would not be due to cross-modal correction.

What would settle it

Take CO2Jump and shuffle the cross-modal attention weights at every step, or replace the CrossSignal with random text confidences, while keeping all remasking counts and schedules identical; if joint accuracy on maze and nonogram solving and image-editing mAP stay about the same, the cross-modal signal is not the cause of the gains.

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

If this is right

  • Joint text-image samples can be produced from a frozen model in one forward pass per step, with no auxiliary evaluator, so concurrent generation need not multiply inference cost.
  • Cross-modal contradictions that previous samplers could not repair — because committed tokens could not be remasked — become correctable during decoding.
  • The reported monotonic improvement with more denoising steps means users can trade compute for joint quality in a predictable way, without the regressions seen in uncoupled baselines.
  • On tasks where text and image are logically interlocked, such as maze and nonogram solving, the two modalities converge to the same algorithmic solution, not merely plausible but mismatched outputs.
  • The chain-rule decomposition provides a principled reason why sharper text grounding should tighten image-side decisions: text decisions are made first within each step, then image scoring is conditioned on them.

Where Pith is reading between the lines

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

  • Because the framework is modality-agnostic, the same self-correcting coupling could be carried over to audio+transcript, code+execution trace, or video+narration, where one stream's late commitments could retract the other's errors; the authors note this as future work.
  • The attention-based surrogate for the missing text state is a testable hypothesis: if it is faithful, then joint NELBO training has implicitly stored cross-modal denoising knowledge in hidden states, suggesting training objectives could be designed to make that knowledge even more accessible at inference.
  • A finer-grained extension would be per-token or per-region entropy gates instead of the single scalar λ; the paper's gating only modulates the global balance between modalities, so local handling of uncertain versus confident regions is left open.
  • The monotonic scaling result hints that the coupling benefit may continue well beyond 512 steps or with adaptive schedules; this is an empirical extrapolation not established by the paper.

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 / 5 minor

Summary. The paper introduces SC-CMJP, a framework for joint text/image generation with masked diffusion models, and CO2Jump, a training-free single-pass sampler that uses cross-modal attention-weighted confidence scores and a remasking jump to couple and self-correct the two modalities. The authors evaluate CO2Jump against MDM, ReMDM, and MMaDA-Parallel on three tasks: image editing with understanding, maze solving, and nonogram solving, using newly curated corpora (JEdit-1M, JMaze-200K, JNono-200K). The central claim is that CO2Jump achieves best joint performance and that its gains compound with more denoising steps because one modality's transition rates are functionals of the other modality's confidence within each step. The paper also positions this as a unified Markov Jump Process formulation with a chain-rule factorization that conditions the image update on the same-step text state.

Significance. If the central mechanism were realized as described, the paper would make a substantial contribution: a training-free, single-forward-pass sampler that couples modalities within a step, with self-correction via remasking, plus large-scale benchmarks with algorithmically verifiable ground truth for maze and nonogram tasks. The empirical setup is extensive and compares against representative baselines, and the datasets could be a useful community resource. However, the paper's own derivation is not implemented by the algorithm: the claimed within-step coupling is replaced by an attention-weighted surrogate that depends only on the previous state, and no rate matrix or target-preservation proof is provided for the asserted Markov Jump Process. The empirical results may be real, but they are currently attributed to a mechanism that is not instantiated.

major comments (4)
  1. [§4.2–4.3, Eq. (7), Algorithm 1] The central chain-rule factorization Eq. (7) says the image update should condition on the text state sampled at the same step, z_text_s. Algorithm 1 does not do this. CoupledConf (Eq. 12) is computed before the text death/birth jumps (lines 10–12 before lines 15–19), and CrossSignal (Eq. 10) uses SelfConf_text extracted from z_t, with the mask-aware bias B_t in Eq. (9) downweighting masked text positions. The text tokens that will be revealed in z_text_s are masked at t and are therefore downweighted or excluded from the image score. Consequently, given z_t, the image update is a deterministic function of z_t (up to Gumbel noise) and is independent of the text update — exactly the uncoupled-parallel failure mode the paper attributes to MMaDA-Parallel in Section 3. The paper acknowledges this approximation in Section 4.2 ('approximating z_text_s with quantities already produced by the si
  2. [§2.2, §4.4] The paper frames SC-CMJP as a Coupled Markov Jump Process with modality-specific rate matrices whose intensities depend on the complementary modality's confidence. No such rate matrix is ever defined. Algorithm 1 is a deterministic top-k/bottom-k selection rule with hard quotas N_remask and N_unmask derived from idealized counts; there are no stochastic jump rates, and no theorem shows that the resulting transition kernel preserves the target distribution. The ReMDM posterior Eq. (4) is a valid modified posterior for a fixed sigma_t, but the score-dependent death/birth selection is not shown to correspond to Eq. (4) or to any Markov kernel. The term 'Markov Jump Process' is therefore not substantiated. Either provide a rate matrix and a proof of marginal preservation, or drop the MJP formalism and describe CO2Jump as a heuristic sampler.
  3. [§4.4, Algorithm 1 lines 16–19] The death and birth quotas are computed from idealized counts U_t^a = floor(alpha_t L_a) and M_t^a = floor((1-alpha_t)L_a), but the actual number of unmasked tokens at runtime can deviate from these idealized values due to previous floor operations, the score-dependent deaths, and the extra-unmask compensation. The paper does not show that the realized token counts follow the intended alpha_t schedule or that the valid-posterior constraint sigma_t <= (1-alpha_s)/alpha_t is satisfied state-by-state. This matters because the ReMDM-style remasking argument in Eq. (4) relies on the schedule being preserved. Please provide a runtime count analysis or explicitly state that CO2Jump intentionally deviates from the idealized schedule and characterize the deviation.
  4. [§5.1, Tables 1–2, Figure 5] The headline image-understanding metric is a COCO-style mAP against a 'pseudoscene graph' constructed by Gemini on the model's own generated target image. No validation of this metric against human annotations or an independent oracle is reported, so absolute mAP values are difficult to interpret. In addition, Table 2 reports no confidence intervals; several improvements over the best baseline are small (e.g., Maze in-distribution joint accuracy: 0.469 vs. 0.461, +0.008), and Figure 5's monotonic-scaling claim is based on a single run without error bars. Please report multiple seeds, confidence intervals, and at least a sanity-check validation of the pseudo-grounding metric.
minor comments (5)
  1. [Abstract/Introduction] Typos: 'geneneration', 'undestanding', 'espectially' should be corrected.
  2. [Algorithm 1, line 12] The comment 'approx. z_text_s via attention' is misleading; the computed quantity is a function of z_text_t, not z_text_s. The comment should say so explicitly.
  3. [Figure 5] The monotonicity claim is strong for a single run; error bars or multiple seeds would make it convincing.
  4. [Table 1] The Qwen3-VL-8B and pre-trained Lumina-DiMOO rows use different numbers of forward passes / turns than the fine-tuned samplers. The text should clarify that these are references, not controlled baselines, to avoid apples-to-oranges reading.
  5. [Figure 7(b)] The caption says 'coupling-driven unmasking' for events in high-Self/low-Cross space, but the surrounding text describes these as remask events; this looks like a typo for 'remasking'.

Circularity Check

0 steps flagged

No significant circularity: CO2Jump is a heuristic inference-time sampler evaluated against external baselines and algorithmic ground truth; the paper's explicit approximation of z_text_s by z_t is an implementation gap, not a definitional identity.

full rationale

The paper's central empirical claim is that CO2Jump outperforms MDM, ReMDM, and MMaDA-Parallel on joint image understanding/editing and on maze/nonogram solving. These results are measured against external baselines and, for the reasoning tasks, against algorithmic ground truth; no reported quantity is a fitted parameter, and no 'prediction' is defined as the value of a fitted input. The theoretical derivation in Eqs. (5)-(12) proposes the chain-rule factorization p(z_text_s, z_image_s | z_t, c) = p(z_text_s | z_t, c) p(z_image_s | z_t, z_text_s, c), and then explicitly approximates the image term by attention-weighted self-confidence from the same forward pass at z_t. The paper states: 'Evaluating Term (II) exactly demands a second forward pass through x_theta at every diffusion step -- prohibitive at inference time. We avoid this cost by approximating z_text_s with quantities already produced by the single forward pass at z_t.' This is an acknowledged approximation, and the skeptic's observation that Algorithm 1 scores images from z_text_t rather than a sampled z_text_s is a correctness/fidelity concern about whether the theory is fully implemented, not a circularity in which the result is equivalent to its inputs by construction. The pseudo-grounding mAP protocol is self-consistent rather than externally grounded, which is a metric-validity concern, but the same protocol is applied to all samplers, and the maze/nonogram results are verified against algorithmic ground truth. I find no load-bearing self-citation: ReMDM, MDM, MMaDA, Lumina-DiMOO, and the oracle MLLMs are external works, and no uniqueness theorem from the authors' prior work is invoked to force the method. The method's reliance on the model's own confidence and attention is a feedback mechanism inherent to sampling, not circular reasoning in the derivation of the empirical claim.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 0 invented entities

The central claim rests on a small set of hand-designed scoring rules and an assumed property of the trained model (that cross-modal attention encodes denoising signal). The free parameters are not fitted to data but are chosen hyperparameters, and the axioms are mostly domain assumptions and ad hoc heuristics rather than proven results. No new physical or conceptual entities are introduced.

free parameters (4)
  • remasking ratio eta / sigma_t schedule = 0.01 on t in [0.25,0.75], 0 elsewhere
    Chosen hyperparameter in experiments for both ReMDM and CO2Jump; not swept or justified, and it controls the amount of remasking.
  • epsilon in entropy gate (Eq. 11) = unspecified small constant
    Used to avoid division by zero; value not reported, affects lambda near zero-entropy regions.
  • mask-aware bias B_t in Eq. (9) = not defined
    Construction is not specified; it is a free design choice that can significantly alter cross-attention weights.
  • number of denoising steps T = swept 8-512 in Figure 5, unspecified for Tables 1-3
    A free hyperparameter; the paper claims monotonic improvement but does not report the value used for headline results.
axioms (5)
  • standard math CTMC equivalence of absorbing-state discrete diffusion (Campbell et al. 2022; Sahoo et al. 2024)
    Used in Section 2.1 to justify the continuous-time interpretation, and in Eq. (4) for the ReMDM posterior.
  • domain assumption The joint NELBO with shared alpha_t factorizes and implicitly forces the network to learn cross-modal denoising signals.
    Section 4.1 asserts that training under Eq. (6) encodes cross-modal clues in hidden representations, but this is assumed, not shown.
  • domain assumption Post-hoc final-layer hidden representations and cross-attention weights are a faithful proxy for the unavailable z_text_s in the chain-rule decomposition.
    Section 4.2/4.3 assumes the single-forward-pass approximation is informative enough; this is the weakest load-bearing assumption.
  • ad hoc to paper Entropy-based gate lambda (Eq. 11) is an appropriate trust measure for combining self and cross signals.
    The formula is introduced heuristically with no derivation or theoretical guarantee; it is a design choice specific to this paper.
  • ad hoc to paper Death and birth quotas based on idealized counts (Algorithm 1) preserve the desired noise schedule while enabling remasking.
    The composition of remask and unmask counts is taken from ReMDM-style reasoning but applied with confidence-based selection; no proof of correctness for this selection rule.

pith-pipeline@v1.3.0-alltime-deepseek · 21637 in / 10065 out tokens · 95932 ms · 2026-08-02T05:54:50.002434+00:00 · methodology

0 comments
read the original abstract

Human cognition does not separate understanding and generation. A teacher at a whiteboard speaks and draws $\textit{together}$, each modality reshapes the other. In this paper, we bring this coupled loop to artificial systems. Masked Diffusion Models (MDMs) are ideally suited to this task, yet existing samplers either decode text and image interleavedly or independently update them in parallel branches that share only previous-step history, but not the other modality's latest decisions $\textit{within}$ the same step; combined with MDMs' inability to remask, cross-modal contradictions are neither detected nor repaired. We introduce $\textbf{Self-Correcting Coupled Markov Jump Processes (SC-CMJP)}$, a framework in which one modality's transition rates are functionals of the other modality's confidence score, as weighted by cross-modal attention. Furthermore, a remasking jump retracts commitments the moment cross-modal evidence turns against them. In conjunction with SC-CMJP, we introduce $\texttt{CO}_\texttt{2}\texttt{Jump}$ (Self-$\underline{\text{CO}}$rrecting $\underline{\text{CO}}$upled $\underline{\text{Jump}}$), a novel training-free single-pass sampler for joint multimodal geneneration. For training and evaluation purposes, we have created and will release three large-scale joint multimodal generation corpora: $\text{JEdit-1M}$, $\text{JMaze-200K}$, $\text{JNono-200K}$, with matching in- and out-of-distribution benchmarks. $\texttt{CO}_\texttt{2}\texttt{Jump}$ achieves best joint performance for image understanding and editing as well as visual reasoning (maze and nonogram solving). The performance of the sampler scales monotonically with the number of denoising steps, evidence that the benefits of cross-modal coupling $\textit{compound}$ across the trajectory. Project page: https://coupled-jump.github.io

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

119 extracted references · 25 linked inside Pith

  1. [1]

    Interpolating Autoregressive and Discrete Denoising Diffusion Language Models , author=

  2. [2]

    Structured denoising diffusion models in discrete state-spaces , author=

  3. [3]

    2025 , eprint=

    Qwen3-VL Technical Report , author=. 2025 , eprint=

    2025

  4. [4]

    A continuous time framework for discrete denoising models , author=

  5. [5]

    Mastering the game of

    Silver, David and Huang, Aja and Maddison, Chris J and Guez, Arthur and Sifre, Laurent and Van Den Driessche, George and Schrittwieser, Julian and Antonoglou, Ioannis and Panneershelvam, Veda and Lanctot, Marc and others , journal=. Mastering the game of. 2016 , publisher=

    2016

  6. [6]

    Discrete flow matching , author=

  7. [7]

    The Journal of chemical physics , volume=

    Approximate accelerated stochastic simulation of chemically reacting systems , author=. The Journal of chemical physics , volume=. 2001 , publisher=

    2001

  8. [8]

    Denoising diffusion probabilistic models , author=

  9. [9]

    2015 , organization=

    Deep unsupervised learning using nonequilibrium thermodynamics , author=. 2015 , organization=

    2015

  10. [10]

    arXiv preprint arXiv:2010.02502 , year=

    Denoising diffusion implicit models , author=. arXiv preprint arXiv:2010.02502 , year=

    Pith/arXiv arXiv 2010

  11. [11]

    arXiv preprint arXiv:2412.10193 , year=

    Simple guidance mechanisms for discrete diffusion models , author=. arXiv preprint arXiv:2412.10193 , year=

    Pith/arXiv arXiv

  12. [12]

    Generative Flows on Discrete State-Spaces: Enabling Multimodal Flows with Applications to Protein Co-Design , author =

  13. [13]

    Unified Diffusion

    Jiayi Chen and Wenxuan Song and Pengxiang Ding and Ziyang Zhou and Han Zhao and Feilong Tang and Donglin Wang and Haoang Li , booktitle=ICLR, year=. Unified Diffusion

  14. [14]

    ArXiv , year=

    SDAR: A Synergistic Diffusion-AutoRegression Paradigm for Scalable Sequence Generation , author=. ArXiv , year=

  15. [15]

    Journal of computational physics , volume=

    A general method for numerically simulating the stochastic time evolution of coupled chemical reactions , author=. Journal of computational physics , volume=. 1976 , publisher=

    1976

  16. [16]

    The journal of physical chemistry , volume=

    Exact stochastic simulation of coupled chemical reactions , author=. The journal of physical chemistry , volume=. 1977 , publisher=

    1977

  17. [17]

    Journal of mathematical physics , volume=

    Time-dependent statistics of the Ising model , author=. Journal of mathematical physics , volume=. 1963 , publisher=

    1963

  18. [18]

    ArXiv , year=

    Don't Settle Too Early: Self-Reflective Remasking for Diffusion Language Models , author=. ArXiv , year=

  19. [19]

    2023 , eprint=

    A Configurable Library for Generating and Manipulating Maze Datasets , author=. 2023 , eprint=

    2023

  20. [20]

    ArXiv , year=

    Fine-Tuning Masked Diffusion for Provable Self-Correction , author=. ArXiv , year=

  21. [21]

    Train for the Worst, Plan for the Best: Understanding Token Ordering in Masked Diffusions , author=

  22. [22]

    Discrete predictor-corrector diffusion models for image synthesis , author=

  23. [23]

    2014 , organization=

    Microsoft coco: Common objects in context , author=. 2014 , organization=

    2014

  24. [24]

    Lou, Aaron and Meng, Chenlin and Ermon, Stefano , title =

  25. [25]

    2025 , url=

    Large Language Diffusion Models , author=. 2025 , url=

    2025

  26. [26]

    2025 , url=

    Your Absorbing Discrete Diffusion Secretly Models the Conditional Distributions of Clean Data , author=. 2025 , url=

    2025

  27. [27]

    2026 , eprint=

    Training-Free Self-Correction for Multimodal Masked Diffusion Models , author=. 2026 , eprint=

    2026

  28. [28]

    ArXiv , year=

    Path Planning for Masked Diffusion Model Sampling , author=. ArXiv , year=

  29. [29]

    OpenAI blog , volume=

    Language models are unsupervised multitask learners , author=. OpenAI blog , volume=

  30. [30]

    Simple and effective masked diffusion language models , author=

  31. [31]

    Simplified and generalized masked diffusion for discrete data , author=

  32. [32]

    Ye Tian and Ling Yang and JiongFan Yang and Anran Wang and Yu Tian and Jiani zheng and Haochen Wang and Zhiyang Teng and Zhuochen Wang and Yinjie Wang and Yunhai Tong and Mengdi Wang and Xiangtai Li , booktitle=ICLR, year=

  33. [33]

    Neural Discrete Representation Learning , author=

  34. [34]

    Generalized Interpolating Discrete Diffusion , author=

  35. [35]

    2025 , url=

    Remasking Discrete Diffusion Models with Inference-Time Scaling , author=. 2025 , url=

    2025

  36. [36]

    2025 , url=

    OmniEdit: Building Image Editing Generalist Models Through Specialist Supervision , author=. 2025 , url=

    2025

  37. [37]

    2025 , url=

    Show-o2: Improved Native Unified Multimodal Models , author=. 2025 , url=

    2025

  38. [38]

    ArXiv , year=

    Lumina-DiMOO: An Omni Diffusion Large Language Model for Multi-Modal Generation and Understanding , author=. ArXiv , year=

  39. [39]

    Ling Yang and Ye Tian and Bowen Li and Xinchen Zhang and Ke Shen and Yunhai Tong and Mengdi Wang , booktitle=NIPS, year=

  40. [40]

    ArXiv , year=

    Dream 7B: Diffusion Large Language Models , author=. ArXiv , year=

  41. [41]

    2025 , url=

    ImgEdit: A Unified Image Editing Dataset and Benchmark , author=. 2025 , url=

    2025

  42. [42]

    2026 , url=

    Informed Correctors for Discrete Diffusion Models , author=. 2026 , url=

    2026

  43. [43]

    Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

    Unifying Continuous and Discrete Text Diffusion with Non-simultaneous Diffusion Processes , author=. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

  44. [44]

    arXiv preprint arXiv:2302.05737 , year=

    A reparameterized discrete diffusion model for text generation , author=. arXiv preprint arXiv:2302.05737 , year=

    Pith/arXiv arXiv

  45. [45]

    arXiv preprint arXiv:2406.01572 , year=

    Unlocking guidance for discrete state-space diffusion and flow models , author=. arXiv preprint arXiv:2406.01572 , year=

    Pith/arXiv arXiv

  46. [46]

    OpenWebText Corpus , author=

  47. [47]

    Attention is all you need , author=

  48. [48]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Scalable diffusion models with transformers , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  49. [49]

    Neurocomputing , volume=

    Roformer: Enhanced transformer with rotary position embedding , author=. Neurocomputing , volume=. 2024 , publisher=

    2024

  50. [50]

    Mauve: Measuring the gap between neural text and human text using divergence frontiers , author=

  51. [51]

    arXiv preprint arXiv:2409.02908 , year=

    Masked diffusion models are secretly time-agnostic masked models and exploit inaccurate categorical sampling , author=. arXiv preprint arXiv:2409.02908 , year=

    Pith/arXiv arXiv

  52. [52]

    arXiv preprint arXiv:1904.09751 , year=

    The curious case of neural text degeneration , author=. arXiv preprint arXiv:1904.09751 , year=

    Pith/arXiv arXiv 1904

  53. [53]

    Divergence frontiers for generative models: Sample complexity, quantization effects, and frontier integrals , author=

  54. [54]

    Variational diffusion models , author=

  55. [55]

    Diffusion models with learned adaptive noise , author=

  56. [56]

    2024 , url=

    Zero-Order Diffusion Guidance for Inverse Problems , author=. 2024 , url=

    2024

  57. [57]

    Diffusion models beat gans on image synthesis , author=

  58. [58]

    2022 , organization=

    GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models , author=. 2022 , organization=

    2022

  59. [59]

    arXiv preprint arXiv:2207.12598 , year=

    Classifier-free diffusion guidance , author=. arXiv preprint arXiv:2207.12598 , year=

    Pith/arXiv arXiv

  60. [60]

    Photorealistic text-to-image diffusion models with deep language understanding , author=

  61. [61]

    2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

    Common Canvas: Open Diffusion Models Trained on Creative-Commons Images , author=. 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=. 2024 , organization=

    2024

  62. [62]

    Generative modeling by estimating gradients of the data distribution , author=

  63. [63]

    Barron and Ben Mildenhall , title =

    Ben Poole and Ajay Jain and Jonathan T. Barron and Ben Mildenhall , title =. The Eleventh International Conference on Learning Representations,. 2023 , url =

    2023

  64. [64]

    2024 , eprint=

    CAT3D: Create Anything in 3D with Multi-View Diffusion Models , author=. 2024 , eprint=

    2024

  65. [65]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Maskgit: Masked generative image transformer , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  66. [66]

    2009 IEEE conference on computer vision and pattern recognition , pages=

    Imagenet: A large-scale hierarchical image database , author=. 2009 IEEE conference on computer vision and pattern recognition , pages=. 2009 , organization=

    2009

  67. [67]

    Gans trained by a two time-scale update rule converge to a local nash equilibrium , author=

  68. [68]

    Journal of chemical information and modeling , volume=

    Enumeration of 166 billion organic small molecules in the chemical universe database GDB-17 , author=. Journal of chemical information and modeling , volume=. 2012 , publisher=

    2012

  69. [69]

    Scientific data , volume=

    Quantum chemistry structures and properties of 134 kilo molecules , author=. Scientific data , volume=. 2014 , publisher=

    2014

  70. [70]

    SMILES, a chemical language and information system. 1. Introduction to methodology and encoding rules , author=. Journal of chemical information and computer sciences , volume=. 1988 , publisher=

    1988

  71. [71]

    ACS central science , volume=

    Molecular transformer: A model for uncertainty-calibrated chemical reaction prediction , author=. ACS central science , volume=. 2019 , publisher=

    2019

  72. [72]

    Greg Landrum , volume=

    RDKit: A software suite for cheminformatics, computational chemistry, and predictive modeling , author=. Greg Landrum , volume=

  73. [73]

    arXiv preprint arXiv:2104.05218 , year=

    FUDGE: Controlled text generation with future discriminators , author=. arXiv preprint arXiv:2104.05218 , year=

    Pith/arXiv arXiv

  74. [74]

    Improved techniques for training gans , author=

  75. [75]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Taming transformers for high-resolution image synthesis , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  76. [76]

    arXiv preprint arXiv:2210.02747 , year=

    Flow matching for generative modeling , author=. arXiv preprint arXiv:2210.02747 , year=

    Pith/arXiv arXiv

  77. [77]

    arXiv preprint arXiv:2011.13456 , year=

    Score-based generative modeling through stochastic differential equations , author=. arXiv preprint arXiv:2011.13456 , year=

    Pith/arXiv arXiv 2011

  78. [78]

    Diffusion-lm improves controllable text generation , author=

  79. [79]

    arXiv preprint arXiv:2210.17432 , year=

    Ssd-lm: Semi-autoregressive simplex-based diffusion language model for text generation and modular control , author=. arXiv preprint arXiv:2210.17432 , year=

    Pith/arXiv arXiv

  80. [80]

    Latent diffusion for language generation , author=

Showing first 80 references.