Pith. sign in

REVIEW 4 major objections 7 minor 57 references

Can Text-to-Image Models Draw from the Right Frame of Reference?

T0 review · 4 major / 7 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Text-to-image models fail when spatial directions refer to an object's own frame of reference, with mean accuracy 41.8% lower than on matched camera-view prompts.

desk verdict A genuinely new T2I evaluation axis with a large, credible deficit, but the exact magnitudes need released artifacts, CIs, and a cleaner prompt contrast before I'd trust the 41.8% headline. read the letter →

arxiv 2608.03357 v1 pith:6TGXSRQR submitted 2026-08-04 cs.CV

Zheyuan Gu , Ruihang Li , Yong Huang , Yiqian Zhang , XIangzhao Hao , Jiaxin Niu , Jiahao Hu , Zhenyu Zhang This is my paper · ORCID
classification cs.CV
keywords frameofreferencetext-to-imagegenerationspatialreasoninglayoutbenchmarkpromptrewritingvision-languageevaluationobjectorientation
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

The paper asks whether text-to-image models can honor an explicitly specified frame of reference when it disagrees with the camera view. To answer it, the authors build FoR-T2I, a benchmark of 1,200 matched prompt pairs: each pair describes the same layout once from the camera view and once through an oriented anchor object's own left, right, front, or back. Across 22 closed- and open-source models, mean final accuracy is 41.8% lower on FoR prompts, and even the strongest model reaches only 44.3% FoR accuracy. The failure concentrates in target placement rather than anchor rendering, and it is sharpest when an anchor-relative left or right maps to the opposite image side, where FoR accuracy collapses to 18.8% against 68.5% on the paired Cam prompts. The paper also shows that a vision-language-model-gated rewriting strategy improves FoR accuracy from 25.0% to 29.2% under the same generation budget.

What carries the argument

The load-bearing object is FoR-T2I's matched prompt-pair construction: each layout deterministically yields one Cam prompt and one FoR prompt that differ only in the frame in which the direction word is resolved. A second mechanism is the frame-mapping trichotomy used in the analysis, which classifies anchor-relative left and right relations as preserved, axis/depth remapped, or reversed relative to the image; the reversed cell is where the FoR deficit becomes a collapse. The automatic evaluator separately scores geometry (target placement), orientation (anchor facing), and their conjunction, which lets the paper attribute the gap to placement rather than orientation.

What would settle it

Re-run the 1,200 layouts with FoR prompts rephrased in short, natural language while keeping the anchor orientation unchanged; if the gap shrinks to near zero, the deficit is largely a phrasing artifact. Alternatively, if any model scores near its Cam accuracy on the reversed-mapping subset, the claim that all models fail reversal would be false.

Watch

Extended reading notes

Core claim

The paper's central claim is that current text-to-image models systematically resolve directional language in the camera's frame and fail to draw from an object's intrinsic frame of reference when the two disagree. This is established by a controlled contrast: the same abstract layout is verbalized as a camera-view prompt and as an object-frame prompt, so any accuracy difference isolates frame-conversion difficulty. The paper reports that every one of 22 models scores lower on FoR prompts; the deficit is greatest for left and right relations under reversal, where the Cam–FoR gap reaches 49.7 points, and component analysis shows the loss appears in target geometry rather than orientation rendering. A secondary claim is that a training-free, vision-language-model-gated prompt-rewriting method can recover a few points of accuracy, but most of the gap remains unresolved.

Load-bearing premise

The Cam–FoR gap is interpreted as frame-conversion difficulty because the paired prompts are meant to differ only in the frame of the direction word, so wording complexity would confound the comparison.

Editorial extensions

If this is right

  • If the claim holds, strong image-frame spatial control does not transfer to object-centered descriptions; the two abilities separate empirically in every model tested.
  • Benchmarks should report frame-of-reference accuracy separately, because aggregate layout accuracy hides the 49.7-point reversal failure.
  • Since the failure survives in the best closed-source models, training-free prompt rewriting alone will not close the gap; orientation-aware representations or training objectives are needed.
  • The vision-language-model-gated rewriting result suggests visual feedback can serve as a cheap partial mitigation in deployed systems without updating the generation model.

Reading between the lines

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

  • An untested extension would be to probe whether the same reversal failure appears in video generation or embodied instruction following, where object-relative 'left' is operationally critical.
  • The 18.8% versus 68.5% reversal finding implies a cheap diagnostic for future models: testing only reversed left/right mappings may predict most of the FoR deficit.
  • Because the benchmark's FoR phrasing is templated and less natural than everyday spatial language, natural rephrasing could change the measured gap; if it does, part of the deficit is a register-matching problem rather than a purely spatial one.
  • A testable prediction is that fine-tuning on reversed-mapping examples with orientation supervision would generalize to unseen anchors and reduce the wrong-frame error pattern.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. The paper introduces FoR-T2I, a benchmark of 1,200 matched camera-view (Cam) and frame-of-reference (FoR) prompt pairs across three task levels, and evaluates 22 closed- and open-source text-to-image models. The central empirical claim is that mean final accuracy is 41.8% lower on FoR prompts than on matched Cam prompts (26.5% vs 15.4%), with the best model reaching only 44.3% FoR accuracy. The paper decomposes the deficit by relation type, camera view, and frame mapping, reporting that the gap is largest when an anchor-relative left/right mapping is reversed relative to the image (49.7 points). It also proposes a training-free VLM-gated rewriting method that improves average FoR accuracy from 25.0% to 29.2% under a matched generation budget.

Significance. If the headline measurement is valid, FoR-T2I is a useful controlled instrument for separating viewer-centered from object-centered spatial instruction following, a distinction existing T2I benchmarks do not isolate. The evaluation is broad (22 models), the Cam-FoR pairing is a sound design, and the paper contains several falsifiable checks: the Level 3 'square of L1' prediction is rejected in the data, the preserved-mapping condition shows only a 5.3-point gap versus 49.7 under reversal, and the literal image-side placement analysis (74.1% vs 16.7%) provides a mechanism for the wrong-frame failure that is largely independent of the orientation judge. These features make the central claim credible and non-circular. The main risk is measurement validity of the automatic evaluator's orientation judgment, which the authors validate only at the aggregate level.

major comments (4)
  1. [§5.5, Table 5b] The automatic evaluator delegates orientation judgment to Qwen3.6-27B, whose standalone agreement with human final-correctness judgments is 65.0%; the full pipeline reaches 87.1%, but no per-condition agreement for Cam vs FoR or for the preserved/remapped/reversed mapping groups is reported. Because FoR scoring requires judging the anchor's rendered orientation, a condition-specific evaluator bias could inflate the 41.8% relative deficit and the 49.7-point reversal gap in §6.2. The literal image-side placement analysis in §6.3 (74.1% vs 16.7%) is good mitigating evidence, but it is not a substitute for per-condition evaluator validation. Please report per-condition human-evaluator agreement and, if feasible, human FoR-only orientation accuracy on the 120-pair validation set, or demonstrate that the main gaps persist under human scoring on that subset.
  2. [§3.2, Figure 3] The construction protocol states that paired prompts 'must differ only in the frame in which the direction word is resolved,' but the displayed FoR prompts are considerably longer and grammatically unnatural (e.g., 'The baristas is behind girl, opposite girl's own facing direction'). While the preserved-mapping control in §6.2 (5.3-point gap) is strong evidence that wording alone does not explain the reversal effect, the paper should explicitly quantify template length and lexical complexity across conditions and state the preserved-mapping result as a control for this confound.
  3. [§5.2, Level 3 analysis] The claim that L3 FoR failure 'costs more than two independent conversions' rests on comparing observed L3 accuracy (10.5%) with the square of L1 accuracy (15.0%). However, the same independence model already overpredicts on Cam prompts (37.5% predicted vs 33.7% observed), so the FoR-specific excess is only about 0.7 points when measured as an absolute shortfall relative to the control condition. Please report the shortfall relative to the Cam baseline (or use a ratio/relative measure) before concluding the excess is specific to FoR.
  4. [Benchmark availability] The paper does not state whether the benchmark data (the 1,200 prompt pairs, the layout engine, and the evaluation code) will be released or where. As the contribution is a benchmark, this is essential for reproducibility and for the field to build on it; please include a clear data/code availability statement and a release plan.
minor comments (7)
  1. [Table 3] The label 'Avg.' is repeated for the geometry block, the orientation block, and the final strict-accuracy columns; the caption should state explicitly which averages are macro-averages over L1-L3 and which columns define final accuracy.
  2. [Figures 1 and 3] There are several typos in displayed prompts: 'choopstics' in Figure 1, and 'A baristas is present' and 'The baristas is behind girl' in Figure 3; please proofread all example prompts.
  3. [References] The reference list contains two entries with identical titles and author lists for 'Wu et al. 2025a' and 'Wu et al. 2025b'; verify that these correspond to distinct models and cite the correct technical report for Qwen-Image-2512.
  4. [Model naming] The model list refers to 'Seedream 5.0 Pro' but the tables use 'Seedream 5.0'; please unify the naming throughout.
  5. [Section 3.2] The text promises that 'Appendix reports our benchmark's statistics,' but no appendix is present in the manuscript; please include the appendix or remove the pointer.
  6. [Section 4] The text states that a language model produces K rewrites but never specifies K or the sampling parameters; please report these for reproducibility.
  7. [Tables 2 and 4] The main results and mitigation results are point estimates without confidence intervals or significance tests; for the central 41.8% relative-deficit claim, at least standard errors on the macro-averaged accuracy would strengthen the paper.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: FoR-T2I's central Cam–FoR gap is an external benchmark measurement, not a quantity derived from its own construction or from self-citation.

full rationale

FoR-T2I makes no derivation claim that reduces to its inputs. The headline 41.8% relative deficit is an external measurement on 1,200 Cam–FoR prompt pairs scored by an automatic pipeline validated against human annotations (Section 5.5), not a quantity defined by a fitted parameter or by the authors' prior results. The L3 'square of L1' prediction (Section 5.2) is a falsifiable independence model, and the paper reports that the data reject it, which is the opposite of circularity. The only self-referential element is that ERNIE-Image, a model from the authors' own ERNIE Team, appears as one of 22 baselines in Table 2 with full disclosure (Section 5.1); this is not load-bearing because the central gap and reversal analysis (Section 6.2) aggregate over all models and hold for closed-source models like GPT-Image-2 and Seedream 5.0. The paired-prompt construction assumption (Section 3.2) that prompts 'must differ only in the frame in which the direction word is resolved' is a benchmark design choice, not a circular definition of the measured outcome. No quoted reduction of a predicted quantity to an input, fitted parameter, or self-citation chain exists.

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

The central claim rests on four assumptions the reader does not pay for upstream: the mapping from the benchmark's two prompt types to Levinson's frame taxonomy, the prompt-isolation assumption that licenses the causal reading of the 41.8% gap, the reliability of the Qwen-family VLM filter that composes the object-orientation pool, and the independence model behind the Level 3 claim. No fitted free parameters appear anywhere: the paper is an empirical measurement, and the L3 squared-accuracy prediction is a parameter-free prediction that the data falsify, which counts as evidence rather than fitting. No invented entities are introduced; FoR-T2I is a constructed measurement artifact, not a postulated physical or conceptual entity with independent existence.

assumptions (4)
  • domain assumption The camera view corresponds to the relative frame and the anchor's own orientation to the intrinsic frame in Levinson's taxonomy; the absolute frame is excluded.
    Section 2 grounds the benchmark in Levinson (2003). If this mapping is wrong, the Cam-FoR contrast does not measure frame-of-reference resolution. Excluding the absolute frame is reasonable for a single image but is still a scope choice that narrows the benchmark's coverage.
  • domain assumption Templated prompts isolate the frame of reference, so the two prompts in a pair differ only in how the direction word is resolved.
    Section 3.2 asserts this identity. The FoR examples in Figure 3 are visibly longer and less natural, so linguistic complexity is a residual confound for the absolute gap; the preserved-versus-reversed comparison in Section 6.2 (5.3 versus 49.7 point gaps) is the main internal support for the assumption.
  • domain assumption The VLM judge used to filter the object-orientation pool (presence, count, orientation consistency) does not bias the pool against the FoR failures being measured.
    Section 3.2 keeps only configurations that pass Qwen-family VLM checks. If the filter shares the evaluated models' orientation biases, the benchmark may under-sample the hardest cases; the direction and size of any bias are not quantified.
  • domain assumption Under the null model, the two frame conversions in Level 3 fail independently, so L3 accuracy should approximate the square of L1 accuracy.
    Section 5.2 predicts L3 FoR accuracy of 15.0% versus an observed 10.5% and treats the shortfall as an excess cost specific to FoR. The independence assumption is untested, and the Cam condition also falls short (37.5 predicted versus 33.7 observed), so the FoR-specific excess is not established.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Can Text-to-Image Models Draw from the Right Frame of Reference?." pith.science (2026). https://pith.science/paper/6TGXSRQR

@misc{pith2026260803357,
  author       = {Pith},
  title        = {Pith review of: Can Text-to-Image Models Draw from the Right Frame of Reference?},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6TGXSRQR}},
  note         = {Machine review of arXiv:2608.03357}
}
read the original abstract

Spatial instruction following has become a crucial requirement for text-to-image (T2I) generation. A common challenge arises when directional expressions are interpreted under different frames of reference. For example, ``the left of'' may refer to the viewer's image coordinates or to the intrinsic orientation of an object, leading to different expected layouts. Existing T2I benchmarks reveal important layout failures, yet they rarely isolate whether models can follow a specified frame of reference when it differs from camera view. To mitigate this gap, we introduce FoR-T2I, a benchmark for evaluating this distinction with 1,200 prompt pairs built from controlled spatial layouts. In each pair, the camera-view (Cam) prompt states the target relation in camera view, while the frame-of-reference (FoR) prompt describes the same target placement through an oriented anchor object. Across 22 closed-source and open-source T2I models, mean final accuracy is 41.8\% lower on FoR prompts than on matched Cam prompts; even the best-performing model achieves only 44.3\% FoR accuracy. This suggests that current models struggle more when the same layout is described through an object's orientation rather than directly in image coordinates. We further analyze this gap by relation type and camera view, compare several training-free prompting and feedback-based mitigation strategies, and propose a VLM-gated rewriting approach that selects rewritten prompts using visual feedback, improving average FoR accuracy from 25.0\% to 29.2\% under the same generation budget.

Figures

Figures reproduced from arXiv: 2608.03357 by the authors.

Figure 1
Figure 1. An example of reference-frame confusion in text [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. FoR-T2I construction pipeline. (1) We extract objects with a well-defined intrinsic front and keep renderings that pass presence, single-instance, and orientation checks. (2) Each layout deterministically specifies the anchor position, target position and anchor orientation. (3) Each layout yields a Camera-View and a FoR prompt that differ only in whether the spatial term is anchored to the camera or to the explicit… view at source ↗
Figure 3
Figure 3. Representative failure patterns under FoR prompts. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 42 canonical work pages

  1. [1]

    , title =

    Levinson, Stephen C. , title =. 2003 , publisher =

    work page 2003

  2. [2]

    2022 , eprint =

    Benchmarking Spatial Relationships in Text-to-Image Generation , author =. 2022 , eprint =

    work page 2022

  3. [3]

    2025 , eprint =

    T2I-CompBench++: An Enhanced and Comprehensive Benchmark for Compositional Text-to-image Generation , author =. 2025 , eprint =

    work page 2025

  4. [4]

    2023 , eprint =

    GenEval: An Object-Focused Framework for Evaluating Text-to-Image Alignment , author =. 2023 , eprint =

    work page 2023

  5. [5]

    2025 , eprint =

    FoREST: Frame of Reference Evaluation in Spatial Reasoning Tasks , author =. 2025 , eprint =

    work page 2025

  6. [6]

    2025 , eprint =

    ViewSpatial-Bench: Evaluating Multi-perspective Spatial Localization in Vision-Language Models , author =. 2025 , eprint =

    work page 2025

  7. [7]

    2025 , eprint =

    GenSpace: Benchmarking Spatially-Aware Image Generation , author =. 2025 , eprint =

    work page 2025

  8. [8]

    2026 , eprint =

    Everything in Its Place: Benchmarking Spatial Intelligence of Text-to-Image Models , author =. 2026 , eprint =

    work page 2026

Show all 57 references
  1. [9]

    2026 , eprint =

    Allocentric Perceiver: Disentangling Allocentric Reasoning from Egocentric Visual Priors via Frame Instantiation , author =. 2026 , eprint =

    2026

  2. [10]

    2026 , eprint =

    Qwen-Image-2.0 Technical Report , author =. 2026 , eprint =

    2026

  3. [11]

    2025 , eprint =

    HunyuanImage 3.0 Technical Report , author =. 2025 , eprint =

    2025

  4. [12]

    2025 , eprint =

    HiDream-I1: A High-Efficient Image Generative Foundation Model with Sparse Diffusion Transformer , author =. 2025 , eprint =

    2025

  5. [13]

    2025 , eprint =

    Z-Image: An Efficient Image Generation Foundation Model with Single-Stream Diffusion Transformer , author =. 2025 , eprint =

    2025

  6. [14]

    2026 , eprint =

    High-Fidelity Two-Step Image Generation via Teacher-Aligned End-to-End Distillation , author =. 2026 , eprint =

    2026

  7. [15]

    2026 , eprint =

    SeFi-Image: A Text-to-Image Foundation Model with Semantic-First Diffusion , author =. 2026 , eprint =

    2026

  8. [16]

    2025 , eprint =

    OmniGen2: Exploration to Advanced Multimodal Generation , author =. 2025 , eprint =

    2025

  9. [17]

    2025 , eprint =

    Emerging Properties in Unified Multimodal Pretraining , author =. 2025 , eprint =

    2025

  10. [18]

    A Survey of Text-to-Image Diffusion Models in Generative AI , year=

    Kandwal, Siddharth and Nehra, Vibha , booktitle=. A Survey of Text-to-Image Diffusion Models in Generative AI , year=

  11. [19]

    2026 , eprint=

    Z-Image: An Efficient Image Generation Foundation Model with Single-Stream Diffusion Transformer , author=. 2026 , eprint=

    2026

  12. [20]

    2026 , eprint=

    SAM 3: Segment Anything with Concepts , author=. 2026 , eprint=

    2026

  13. [21]

    International Conference on Learning Representations (ICLR) , year=

    VisualPrompter: Semantic-Aware Prompt Optimization with Visual Feedback for Text-to-Image Synthesis , author=. International Conference on Learning Representations (ICLR) , year=

  14. [22]

    2025 , eprint =

    VisionDirector: Vision-Language Guided Closed-Loop Refinement for Generative Image Synthesis , author =. 2025 , eprint =

    2025

  15. [23]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year=

    T2I-Copilot: A Training-Free Multi-Agent Text-to-Image System for Enhanced Prompt Interpretation and Interactive Generation , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year=

  16. [24]

    ICLR , year =

    Jaemin Cho and Yushi Hu and Roopal Garg and Peter Anderson and Ranjay Krishna and Jason Baldridge and Mohit Bansal and Jordi Pont-Tuset and Su Wang , title =. ICLR , year =

  17. [25]

    arXiv preprint arXiv:2303.11897 , year=

    TIFA: Accurate and Interpretable Text-to-Image Faithfulness Evaluation with Question Answering , author=. arXiv preprint arXiv:2303.11897 , year=

    arXiv

  18. [26]

    2025 , howpublished =

    GPT Image 1 , author =. 2025 , howpublished =

    2025

  19. [27]

    2026 , eprint=

    TIIF-Bench: How Does Your T2I Model Follow Your Instructions? , author=. 2026 , eprint=

    2026

  20. [28]

    arXiv preprint arxiv:2506.07977 , year=

    OneIG-Bench: Omni-dimensional Nuanced Evaluation for Image Generation , author=. arXiv preprint arxiv:2506.07977 , year=

    arXiv

  21. [29]

    arXiv preprint arXiv:2511.10647 , year =

    Depth Anything 3: Recovering the Visual Space from Any Views , author =. arXiv preprint arXiv:2511.10647 , year =

    arXiv

  22. [30]

    2026 , eprint =

    SAM 3: Segment Anything with Concepts , author =. 2026 , eprint =

    2026

  23. [31]

    2026 , howpublished =

    GPT Image 2 , author =. 2026 , howpublished =

    2026

  24. [32]

    2025 , howpublished =

    Gemini 3 Pro Image , author =. 2025 , howpublished =

    2025

  25. [33]

    2026 , howpublished =

    Gemini 3.1 Flash Image , author =. 2026 , howpublished =

    2026

  26. [34]

    2025 , howpublished =

    Gemini 2.5 Flash Image , author =. 2025 , howpublished =

    2025

  27. [35]

    2025 , howpublished =

    Imagen 4 , author =. 2025 , howpublished =

    2025

  28. [36]

    2026 , howpublished =

    Seedream 5.0 Pro , author =. 2026 , howpublished =

    2026

  29. [37]

    arXiv preprint arXiv:2509.20427 , year =

    Seedream 4.0: Toward Next-generation Multimodal Image Generation , author =. arXiv preprint arXiv:2509.20427 , year =

    arXiv

  30. [38]

    2025 , eprint =

    Qwen-Image Technical Report , author =. 2025 , eprint =

    2025

  31. [39]

    2026 , howpublished =

    ERNIE-Image , author =. 2026 , howpublished =

    2026

  32. [40]

    2025 , eprint =

    Skywork UniPic 2.0: Building Kontext Model with Online RL for Unified Multimodal Model , author =. 2025 , eprint =

    2025

  33. [41]

    2024 , eprint =

    Infinity: Scaling Bitwise AutoRegressive Modeling for High-Resolution Image Synthesis , author =. 2024 , eprint =

    2024

  34. [42]

    arXiv preprint arXiv:2501.17811 , year =

    Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling , author =. arXiv preprint arXiv:2501.17811 , year =

    arXiv

  35. [43]

    arXiv preprint arXiv:2408.12528 , year =

    Show-o: One Single Transformer to Unify Multimodal Understanding and Generation , author =. arXiv preprint arXiv:2408.12528 , year =

    arXiv

  36. [44]

    2024 , eprint =

    Scaling Rectified Flow Transformers for High-Resolution Image Synthesis , author =. 2024 , eprint =

    2024

  37. [45]

    2014 , eprint=

    Generative Adversarial Networks , author=. 2014 , eprint=

    2014

  38. [46]

    2022 , eprint=

    High-Resolution Image Synthesis with Latent Diffusion Models , author=. 2022 , eprint=

    2022

  39. [47]

    2023 , eprint=

    Flow Matching for Generative Modeling , author=. 2023 , eprint=

    2023

  40. [48]

    Journal of Memory and Language , volume =

    Perspective in Spatial Descriptions , author =. Journal of Memory and Language , volume =. 1996 , doi =

    1996

  41. [49]

    Journal of Pragmatics , volume =

    Reference Frames of Space and Time in Language , author =. Journal of Pragmatics , volume =. 2011 , doi =

    2011

  42. [50]

    Journal of Memory and Language , volume =

    The Influence of Reference Frame Selection on Spatial Template Construction , author =. Journal of Memory and Language , volume =. 1997 , doi =

    1997

  43. [51]

    2023 , eprint =

    Text-to-image Diffusion Models in Generative AI: A Survey , author =. 2023 , eprint =

    2023

  44. [52]

    ACM Computing Surveys , volume =

    Diffusion Models: A Comprehensive Survey of Methods and Applications , author =. ACM Computing Surveys , volume =. 2023 , doi =

    2023

  45. [53]

    2024 , eprint =

    Controllable Generation with Text-to-Image Diffusion Models: A Survey , author =. 2024 , eprint =

    2024

  46. [54]

    International Journal of Computer Vision , year =

    Diffusion Models in Robotics: A Survey , author =. International Journal of Computer Vision , year =

  47. [55]

    ACM Computing Surveys , volume =

    Understanding World or Predicting Future? A Comprehensive Survey of World Models , author =. ACM Computing Surveys , volume =

  48. [56]

    2025 , eprint=

    Qwen-Image Technical Report , author=. 2025 , eprint=

    2025

  49. [57]

    2025 , eprint=

    SAM 3: Segment Anything with Concepts , author=. 2025 , eprint=

    2025

Pith tools

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