REVIEW 5 major objections 8 minor 54 references
CoherenDream: Boosting Holistic Text Coherence in 3D Generation via Multimodal Large Language Models Feedback
T0 review · 5 major / 8 minor · reviewed 2026-08-16 · deepseek-v4-flash
Pith's one-line read Feeding multimodal-LLM text-alignment feedback into score distillation as a correction term fixes object omissions and relational errors in text-to-3D generation.
desk verdict Real novelty in MLLM-guided SDS for 3D, with solid ablations; but the gradient derivation is heuristic and the evaluation is too thin for the claims—still deserves review. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing object is the textual-consistent noise prediction $\hat{\epsilon}_\Phi=\epsilon_\Phi+\lambda\,\delta(T,\hat{x}_0)$, where $\delta(T,\hat{x}_0)=\nabla_{x_t}L_{\mathrm{ce}}(T,f_{\mathrm{cr}}(\hat{x}_0))$ is the gradient of an MLLM critic's cross-entropy loss against prompt-derived targets on a one-step-denoised rendering. It turns a language-model judgment of which objects, relations, and views are present into a differentiable correction to the SDS update direction. The critic is decomposed into three question-answering tasks (scene graph generation, multi-label object classification, and view classification) so the feedback can target global semantics, missing objects, and geometry separately. The auxiliary LLM-layout initialization plays a supporting role by fitting the density field's occupancy to a rough box layout during the first hundreds of steps, so the diffusion prior starts from a semantically plausible arrangement.
What would settle it
Take a fixed 3D scene whose prompt and renderings are deliberately mismatched, for example a prompt naming two macaws playing chess but a scene containing one macaw and no chessboard; one-step denoise renderings across a range of noise levels and check whether the TCSD correction $\delta(T,\hat{x}_0)$ points in the same direction as a trusted alignment score or human judgment. Systematic sign disagreement at high noise would show the feedback proxy is unreliable there.
Extended reading notes
Core claim
The central claim is that the accumulated view-independent bias of SDS can be corrected online by an MLLM that reads the current renderings and reports how faithfully they answer prompt-derived questions. Writing the ideal distribution as $\hat{p}_t(x_t\mid y)=p_t(x_t\mid y)\,p(y\mid x_t)$ and using the score-function identity $\nabla_{x_t}\log p_t(x_t\mid y)=-\epsilon_\Phi/\sqrt{1-\alpha_t}$, the paper replaces the unknown term $\nabla_{x_t}\log p(y\mid x_t)$ with $\lambda\,\nabla_{x_t}L_{\mathrm{ce}}(T,f_{\mathrm{cr}}(\hat{x}_0))$ evaluated on the one-step denoising $\hat{x}_0$, yielding the corrected noise prediction $\hat{\epsilon}_\Phi=\epsilon_\Phi+\lambda\,\delta(T,\hat{x}_0)$ in Eq. (7). The correction is computed for three guidance tasks: scene graph generation, multi-label object classification, and view classification, and it is what lets the optimization steer toward a distribution in which all named objects and their relations appear in a holistic three-dimensional representation. The paper further claims that fine-tuning the critic on view-aware, diffusion-noisy grid images and initializing the scene from an LLM-generated layout make this feedback accurate enough to produce text-coherent results that a vanilla SDS baseline misses.
Load-bearing premise
The correction term assumes the fine-tuned MLLM's cross-entropy loss on one-step-denoised renderings is a faithful, properly scaled differentiable proxy for how well the prompt matches the image, so a miscalibrated critic could push the 3D scene away from user intent.
Editorial extensions
If this is right
- TCSD can be applied on top of existing SDS variants: integrating it with JointDreamer fixes omitted objects, and combining it with variational score distillation adds texture fidelity while keeping prompt alignment.
- Because the correction acts on the noise prediction rather than replacing the diffusion model, it preserves the base model's generalization and avoids fine-tuning a diffusion model on 3D datasets.
- The three-task feedback decomposition gives a practical recipe for using MLLMs as optimizers, not just evaluators, in generative 3D pipelines.
- Using TCSD with a holistic NeRF representation yields more faithful object interactions than compositional methods that optimize objects separately.
- CoherenDream's training time stays comparable to the baseline (about 60 minutes), so the semantic feedback adds little computational overhead.
Reading between the lines
- The TCSD correction is effectively a differentiable semantic reward, so the same one-step-denoising trick could transfer to image-to-3D or test-time prompt-adherence settings where a critic can be queried on intermediate outputs.
- Because $\hat{x}_0$ gets blurrier at high noise levels, the reliability of $\delta$ should degrade there; annealing $\lambda$ by noise level or scheduling which guidance tasks fire at which timesteps is a natural testable improvement.
- Comparing the direction of the TCSD correction with a trusted reference alignment signal (such as human preference ratings) would diagnose how much the critic's loss can be trusted; the paper's own failure cases show attribute leakage when prompt supervision is weak.
- The view-aware data collection pipeline consumes only rendered grids, so the same critic recipe could be reused for Gaussian-splatting or mesh-based 3D generators.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proposes CoherenDream, a text-to-3D generation framework that augments Score Distillation Sampling (SDS) with feedback from a multimodal large language model (MLLM). The central contribution is Textual Coherent Score Distillation (TCSD), which modifies the SDS noise prediction as \hat{\epsilon}_\Phi = \epsilon_\Phi + \lambda\cdot\delta(T,\hat{x}_0), where \delta is the gradient of an MLLM cross-entropy loss evaluated on one-step denoised renderings against scene-graph, object, and view targets. The authors fine-tune a 3DLLaVA-CRITIC model on a view-aware dataset generated with GPT-4o, and they introduce an LLM-layout initialization that warms up the NeRF density field. Experiments on a 45-prompt TIFA subset report improved TIFA, VQAScore, and CLIP scores over MVDream, DreamView, and JointDreamer, with additional ablations, comparisons to compositional methods and native 3D models, and integration experiments with JSD and VSD.
Significance. If the central mechanism is validated, integrating MLLM feedback directly into SDS is a promising and timely direction that could mitigate the view-independent bias accumulation in multi-object text-to-3D generation. The paper appears to be the first to embed an MLLM in the SDS optimization loop, and the view-aware data collection pipeline, the decomposition of alignment feedback into scene-graph, object, and view tasks, and the extensive qualitative comparisons are useful contributions. The compatibility of TCSD with JSD and VSD demonstrates that the approach is modular and could be adopted by other SDS variants. However, the current evidence does not establish that the added gradient is a faithful alignment-score correction: the derivation is heuristic, the critic validation is circular, and the headline quantitative evaluation is based on a small prompt subset with nonstandard aggregation and no error bars. The paper's own failure cases further show that the MLLM feedback can inject attribute biases, so the central claim requires additional calibration and controlled experiments.
major comments (5)
- [Textual Coherent Score Distillation (TCSD), Eqs. (4)-(7)] The substitution of \nabla_{\mathbf{x}_t}\log p(y|\mathbf{x}_t) by \lambda\cdot\nabla_{\mathbf{x}_t} L_{\mathrm{ce}}(T, f_{\mathrm{cr}}(\hat{\mathbf{x}}_0)) is asserted rather than derived or calibrated. The cross-entropy of an autoregressive MLLM is a token-prediction loss on 2x2 grid images, not a calibrated estimate of the probability that \mathbf{x}_t is aligned with the user prompt. Moreover, \hat{\mathbf{x}}_0 is obtained by one-step denoising, whereas the critic's training images were generated with 5/20/50 denoising steps (Appendix, "View-aware Image Generation Details"), so at the large timesteps used early in SDS the critic operates far outside its training distribution. I request either a derivation of this replacement with error bounds, or a calibration experiment showing that L_{\mathrm{ce}} correlates with human-judged text alignment on SDS-like noisy renderings, and a control experiment comparing optimization trajectories with and without the MLLM term.
- [Quantitative Comparison, Table 1] The headline improvement is based on a curated 45-prompt subset of TIFA v1.0 with no error bars, confidence intervals, or significance tests. The aggregation rules are nonstandard and differ across metrics: TIFA is the intersection of correct answers across 10 uniformly sampled azimuths, VQAScore is the maximum across viewpoints, and CLIP is the average across views. These choices can systematically favor different failure modes; for example, max-aggregated VQAScore rewards methods that occasionally produce a single good view, while intersection-based TIFA penalizes view inconsistency. The paper should justify these aggregation choices, report per-prompt scores and variance, and ideally evaluate on the full TIFA set or another standardized benchmark to establish that the observed 4.0-point TIFA improvement is not an artifact of prompt selection or aggregation.
- [Quantitative evaluation for 3DLLaVA-CRITIC, Table 2 and Appendix "Quality of 3DLLaVA-CRITIC"] The validation of the critic is circular: GPT-4o generated the fine-tuning labels in the view-aware data collection pipeline, and GPT-4o also scores the critic's responses as correct or incorrect in the reported validation. The accuracies (T_sg 0.76, T_ob 0.89, T_view 0.87) therefore measure agreement with the same labeler that created the training data, not true accuracy on an independent ground truth. Since the critic's feedback directly steers the SDS gradient, an independent human evaluation on a sampled subset, or a held-out set annotated by a different source, is necessary to trust the TCSD results and to determine whether the critic's errors introduce biases into the generated 3D assets.
- [Guidance Task Combo, Eq. (7)] The task weights \lambda_{\mathrm{sg}}=1, \lambda_{\mathrm{ob}}=0.5, and \lambda_{\mathrm{view}}=0.1 are fixed without a sensitivity analysis or any comparison of gradient magnitudes between the SDS noise term \epsilon_\Phi and the MLLM feedback term \lambda\cdot\delta. The paper also notes that view classification slows training and is therefore computed only every 20 iterations, which changes the effective weight of that task across time. Without ablations that vary these weights, or at least report the norm of the feedback gradient relative to the SDS gradient during optimization, it is unclear whether the MLLM term dominates, is negligible, or is balanced at different phases of training; the reported improvements could be highly sensitive to these ad-hoc values.
- [Limitation and Fig. 19] The acknowledged failure cases, attribute leakage (the "green" from the stool influencing the monkey) and missed attributes (the pig's "reclined position" not rendered), show that the MLLM feedback can inject biases rather than consistently correct view-independent drift. This directly qualifies the central claim that TCSD "steers the optimization toward a textual-consistent distribution." The paper should analyze why these biases arise, for example whether the scene-graph format over-weights object presence and under-weights attributes and relations, and should propose a mitigation or at least quantify the frequency of such failures across a larger prompt set.
minor comments (8)
- [Baselines Setup] In the Appendix, "1DreamView" should be "DreamView", and in the qualitative comparison section "our Conherent" should be "our CoherenDream".
- [Framework of CoherenDream] In "Guidance Tasks", "equipe" should be "equip", and in "LLM-layout Initialization", "traption" should be "trapping"; in Fig. 6, "TCDC" should be "TCSD".
- [Preliminaries, Eq. (2)] The notation E_{t,\epsilon_\Phi} is ambiguous because the sampling distributions of t and \epsilon are not defined; please specify the noise schedule, the range of t, and the distribution of \epsilon.
- [Eq. (5) and Guidance Task Combo] The symbol \lambda is used both for \sqrt{1-\alpha_t} in Eq. (5) and for the task weights \lambda_{\mathrm{sg}}, \lambda_{\mathrm{ob}}, \lambda_{\mathrm{view}}; please use a distinct symbol for the task weights to avoid confusion.
- [References and notation] The model name appears variously as "3DLLaV A-CRITIC", "3DLLaVA-CRITIC", and "3DLlava-CRITIC"; please standardize, and the reference list entry for LLaVA-OneVision is incomplete.
- [Fig. 16, layout prompt] The example JSON in the layout generation prompt contains syntax errors, including double braces and a trailing comma, which would not parse; please fix the example.
- [Quantitative Comparison] The paper refers to a "curated 45-prompt subset" of TIFA v1.0 but does not specify the curation procedure or list the prompts; adding a supplementary prompt list would aid reproducibility.
- [Table 4 in Appendix] The heading "Quantity ablation study" should be "Quantitative ablation study", and the object classification prompt in Table 3 lacks an output format example, making it hard to reproduce the model's structured output.
Circularity Check
Main TIFA/CLIP evaluation is external and not circular; only the Table 2 validation of 3DLLaVA-CRITIC is self-referential, with GPT-4o serving as both labeler and judge.
-
fitted input called prediction
[Section 3DLLaVA-CRITIC / quantitative evaluation of 3DLLaVA-CRITIC (Table 2)]
"In the absence of a ground-truth dataset, we construct 30 prompts for T_sg and T_ob, and randomly choose 10 objects across 10 camera poses for T_view. For each validation sample, GPT-4o is given both models’ responses alongside the reference answer and tasked with scoring each as correct (1) or incorrect (0)."
The reference answers in this validation come from the same GPT-4o annotation step that produced the critic's instruction-tuning data ('we curate 307,409 grid images and then prompt GPT-4o ... to generate scene graphs, view prompts, and perform object extraction'). GPT-4o is then asked to judge whether the fine-tuned critic matches those references. The resulting Table 2 'accuracy' is therefore agreement with the teacher that generated the labels, not an independent measure of critic quality. This is a narrow evaluation of the auxiliary critic, however, and does not by itself establish the paper's central text-to-3D claim, which is supported by external TIFA/CLIP/VQAScore numbers.
full rationale
The central derivation is not circular: TCSD takes the standard SDS gradient and adds an MLLM-feedback correction term. Equation (5)'s replacement of ∇_x_t log p(y|x_t) by λ·∇_x_t Lce(T, f_cr(x̂_0)) is an unverified approximation rather than a definitional equivalence; it may be a correctness or calibration risk, but it does not reduce the prediction to the input. The headline results in Table 1 (TIFA, VQAScore, CLIP) are evaluated with external benchmarks and are not constructed from the method's fitted values. The LLM-layout initialization is an initialization heuristic, not a fitted parameter disguised as a prediction. The only identifiable circular element is the Table 2 validation of 3DLLaVA-CRITIC, where GPT-4o both generated the training references and scores the validation answers; that makes the reported critic accuracy an agreement-with-teacher measure. Since this affects an auxiliary component and not the main external evaluation, the overall circularity score is low.
Assumptions & free parameters
free parameters (4)
- 3DLLaVA-CRITIC fine-tuned weights (LoRA and multimodal MLP adapter) =
Trained on 307,409 GPT-4o-annotated grid images, LLaVA-OneVision 0.5B base
- lambda_sg, lambda_ob, lambda_view task weights =
1.0, 0.5, 0.1
- sigma (surface weight decay hyperparameter in LLM-layout loss) =
Not reported
- Warm-up schedule N and layout loss weight decay =
600 steps, decay from 1 to 1e-3
assumptions (5)
- standard math Score identity gradient of log p_t(x_t|y) equals -1/sqrt(1-alpha_t) times the noise prediction (standard diffusion score matching).
- standard math SDS gradient ignores the UNet Jacobian, approximating the KL gradient by the noise-prediction residual.
- ad hoc to paper MLLM cross-entropy L_ce(T, f_cr(hat_x0)) is a faithful differentiable proxy for -log p(y|x_t).
- domain assumption One-step denoising approximates the clean image for feedback computation.
- domain assumption GPT-4-generated LLM layouts provide plausible spatial priors for initialization.
Cite this review
Pith. "Pith review of CoherenDream: Boosting Holistic Text Coherence in 3D Generation via Multimodal Large Language Models Feedback." pith.science (2026). https://pith.science/paper/AL7NYIYH
@misc{pith2026250419860,
author = {Pith},
title = {Pith review of: CoherenDream: Boosting Holistic Text Coherence in 3D Generation via Multimodal Large Language Models Feedback},
year = {2026},
howpublished = {\url{https://pith.science/paper/AL7NYIYH}},
note = {Machine review of arXiv:2504.19860}
}
read the original abstract
Score Distillation Sampling (SDS) has achieved remarkable success in text-to-3D content generation. However, SDS-based methods struggle to maintain semantic fidelity for user prompts, particularly when involving multiple objects with intricate interactions. While existing approaches often address 3D consistency through multiview diffusion model fine-tuning on 3D datasets, this strategy inadvertently exacerbates text-3D alignment degradation. The limitation stems from SDS's inherent accumulation of view-independent biases during optimization, which progressively diverges from the ideal text alignment direction. To alleviate this limitation, we propose a novel SDS objective, dubbed as Textual Coherent Score Distillation (TCSD), which integrates alignment feedback from multimodal large language models (MLLMs). Our TCSD leverages cross-modal understanding capabilities of MLLMs to assess and guide the text-3D correspondence during the optimization. We further develop 3DLLaVA-CRITIC - a fine-tuned MLLM specialized for evaluating multiview text alignment in 3D generations. Additionally, we introduce an LLM-layout initialization that significantly accelerates optimization convergence through semantic-aware spatial configuration. Our framework, CoherenDream, achieves consistent improvement across multiple metrics on TIFA subset.As the first study to incorporate MLLMs into SDS optimization, we also conduct extensive ablation studies to explore optimal MLLM adaptations for 3D generation tasks.
Figures
Figures from the paper (16 more)
Reference graph
Works this paper leans on
-
[1]
, " * write output.state after.block = add.period write newline
ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...
-
[2]
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 word.in bbl.in capitalize " " * FUNCT...
-
[3]
Anil, R.; Dai, A. M.; Firat, O.; Johnson, M.; Lepikhin, D.; Passos, A.; Shakeri, S.; Taropa, E.; Bailey, P.; Chen, Z.; et al. 2023. Palm 2 technical report. arXiv preprint arXiv:2305.10403
arXiv 2023
-
[4]
Armandpour, M.; Zheng, H.; Sadeghian, A.; Sadeghian, A.; and Zhou, M. 2024. Re-imagine the Negative Prompt Algorithm: Transform 2D Diffusion into 3D, alleviate Janus problem and Beyond. In ICLR
work page 2024
-
[5]
Bai, H.; Lyu, Y.; Jiang, L.; Li, S.; Lu, H.; Lin, X.; and Wang, L. 2023. CompoNeRF: Text-guided multi-object compositional NeRF with editable 3D scene layout. arXiv preprint arXiv:2303.13843
arXiv 2023
-
[6]
Chen, R.; Chen, Y.; Jiao, N.; and Jia, K. 2023. Fantasia3d: Disentangling geometry and appearance for high-quality text-to-3d content creation. In ICCV, 22246--22256
work page 2023
-
[7]
Cho, J.; Zala, A.; and Bansal, M. 2023. Visual programming for step-by-step text-to-image generation and evaluation. Advances in Neural Information Processing Systems, 36: 6048--6069
work page 2023
-
[8]
Fang, Y.; Sun, Z.; Wu, T.; Wang, J.; Liu, Z.; Wetzstein, G.; and Lin, D. 2024. Make-it-Real: Unleashing Large Multimodal Model's Ability for Painting 3D Objects with Realistic Materials. arXiv preprint arXiv:2404.16829
arXiv 2024
Show all 54 references
-
[9]
E.; and Wang, W
Feng, W.; Zhu, W.; Fu, T.-j.; Jampani, V.; Akula, A.; He, X.; Basu, S.; Wang, X. E.; and Wang, W. Y. 2024 a . Layoutgpt: Compositional visual planning and generation with large language models. Advances in Neural Information Processing Systems, 36
2024
-
[10]
Feng, Y.; Gong, B.; Chen, D.; Shen, Y.; Liu, Y.; and Zhou, J. 2024 b . Ranni: Taming text-to-image diffusion for accurate instruction following. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 4744--4753
2024
-
[11]
Gao, G.; Liu, W.; Chen, A.; Geiger, A.; and Sch \"o lkopf, B. 2024. Graphdreamer: Compositional 3d scene synthesis from scene graphs. In CVPR, 21295--21304
2024
-
[12]
Guo, Y.-C.; Liu, Y.-T.; Shao, R.; Laforte, C.; Voleti, V.; Luo, G.; Chen, C.-H.; Zou, Z.-X.; Wang, C.; Cao, Y.-P.; and Zhang, S.-H. 2023. threestudio: A unified framework for 3D content generation. https://github.com/threestudio-project/threestudio
2023
-
[13]
He, Y.; Bai, Y.; Lin, M.; Zhao, W.; Hu, Y.; Sheng, J.; Yi, R.; Li, J.; and Liu, Y.-J. 2023. T3Bench: Benchmarking Current Progress in Text-to-3D Generation. arXiv preprint arXiv:2310.02977
2023 arXiv
-
[14]
L.; and Choi, Y
Hessel, J.; Holtzman, A.; Forbes, M.; Bras, R. L.; and Choi, Y. 2021. CLIPScore: A Reference-free Evaluation Metric for Image Captioning. In EMNLP
2021
-
[15]
J.; Shen, Y.; Wallis, P.; Allen-Zhu, Z.; Li, Y.; Wang, S.; Wang, L.; and Chen, W
Hu, E. J.; Shen, Y.; Wallis, P.; Allen-Zhu, Z.; Li, Y.; Wang, S.; Wang, L.; and Chen, W. 2021. Lora: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685
2021 arXiv
-
[16]
Hu, X.; Wang, R.; Fang, Y.; Fu, B.; Cheng, P.; and Yu, G. 2024. Ella: Equip diffusion models with llm for enhanced semantic alignment. arXiv preprint arXiv:2403.05135
2024 arXiv
-
[17]
Hu, Y.; Liu, B.; Kasai, J.; Wang, Y.; Ostendorf, M.; Krishna, R.; and Smith, N. A. 2023. Tifa: Accurate and interpretable text-to-image faithfulness evaluation with question answering. In ICCV
2023
-
[18]
Huang, Y.; Wang, J.; Shi, Y.; Tang, B.; Qi, X.; and Zhang, L. 2023. Dreamtime: An improved optimization strategy for diffusion-guided 3d generation. In ICLR
2023
-
[19]
Jiang, C.; Zeng, Y.; Hu, T.; Xu, S.; Zhang, W.; Xu, H.; and Yeung, D.-Y. 2025. JointDreamer: Ensuring Geometry Consistency and Text Congruence in Text-to-3D Generation via Joint Score Distillation. In ECCV
2025
-
[20]
Kirstain, Y.; Polyak, A.; Singer, U.; Matiana, S.; Penna, J.; and Levy, O. 2023. Pick-a-pic: An open dataset of user preferences for text-to-image generation. Advances in Neural Information Processing Systems, 36: 36652--36663
2023
-
[21]
Li, B.; Zhang, Y.; Guo, D.; Zhang, R.; Li, F.; Zhang, H.; Zhang, K.; Li, Y.; Liu, Z.; and Li, C. 2024 a . Llava-onevision: Easy visual task transfer. arXiv preprint arXiv:2408.03326
2024 arXiv
-
[22]
Li, W.; Chen, R.; Chen, X.; and Tan, P. 2024 b . SweetDreamer: Aligning Geometric Priors in 2D Diffusion for Consistent Text-to-3D. In ICLR
2024
-
[23]
Li, W.; Liu, J.; Yan, H.; Chen, R.; Liang, Y.; Chen, X.; Tan, P.; and Long, X. 2024 c . Craftsman3d: High-fidelity mesh generation with 3d native generation and interactive geometry refiner. In CVPR
2024
-
[24]
Lian, L.; Li, B.; Yala, A.; and Darrell, T. 2023. Llm-grounded diffusion: Enhancing prompt understanding of text-to-image diffusion models with large language models. arXiv preprint arXiv:2305.13655
2023 arXiv
-
[25]
Liang, Y.; Yang, X.; Lin, J.; Li, H.; Xu, X.; and Chen, Y. 2024. Luciddreamer: Towards high-fidelity text-to-3d generation via interval score matching. In CVPR
2024
-
[26]
Lin, C.-H.; Gao, J.; Tang, L.; Takikawa, T.; Zeng, X.; Huang, X.; Kreis, K.; Fidler, S.; Liu, M.-Y.; and Lin, T.-Y. 2023 a . Magic3D: High-Resolution Text-to-3D Content Creation. In CVPR
2023
-
[27]
Lin, Y.; Wu, H.; Wang, R.; Lu, H.; Lin, X.; Xiong, H.; and Wang, L. 2023 b . Towards language-guided interactive 3d generation: Llms as layout interpreter with generative feedback. arXiv preprint arXiv:2305.15808
2023 arXiv
-
[28]
Lin, Z.; Pathak, D.; Li, B.; Li, J.; Xia, X.; Neubig, G.; Zhang, P.; and Ramanan, D. 2024. Evaluating text-to-visual generation with image-to-text generation. In ECCV
2024
-
[29]
Liu, Y.; Lin, C.; Zeng, Z.; Long, X.; Liu, L.; Komura, T.; and Wang, W. 2024. Syncdreamer: Generating multiview-consistent images from a single-view image. In ICLR
2024
-
[30]
Luo, W.; Hu, T.; Zhang, S.; Sun, J.; Li, Z.; and Zhang, Z. 2024. Diff-instruct: A universal approach for transferring knowledge from pre-trained diffusion models. NeurIPS, 36
2024
-
[31]
Metzer, G.; Richardson, E.; Patashnik, O.; Giryes, R.; and Cohen-Or, D. 2023. Latent-nerf for shape-guided generation of 3d shapes and textures. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 12663--12673
2023
-
[32]
P.; Tancik, M.; Barron, J
Mildenhall, B.; Srinivasan, P. P.; Tancik, M.; Barron, J. T.; Ramamoorthi, R.; and Ng, R. 2021. Nerf: Representing scenes as neural radiance fields for view synthesis. Communications of the ACM, 65(1): 99--106
2021
-
[33]
M \"u ller, T.; Evans, A.; Schied, C.; and Keller, A. 2022. Instant neural graphics primitives with a multiresolution hash encoding. ACM Transactions on Graphics (ToG), 41(4): 1--15
2022
-
[34]
OpenAI, R. 2023. Gpt-4 technical report. arxiv 2303.08774. View in Article, 2(5)
2023 arXiv
-
[35]
T.; and Mildenhall, B
Poole, B.; Jain, A.; Barron, J. T.; and Mildenhall, B. 2023. Dreamfusion: Text-to-3d using 2d diffusion. In ICLR
2023
-
[36]
Qiu, L.; Chen, G.; Gu, X.; Zuo, Q.; Xu, M.; Wu, Y.; Yuan, W.; Dong, Z.; Bo, L.; and Han, X. 2024. Richdreamer: A generalizable normal-depth diffusion model for detail richness in text-to-3d. In CVPR
2024
-
[37]
Rombach, R.; Blattmann, A.; Lorenz, D.; Esser, P.; and Ommer, B. 2022. High-resolution image synthesis with latent diffusion models. In CVPR, 10684--10695
2022
-
[38]
Seo, J.; Jang, W.; Kwak, M.-S.; Kim, H.; Ko, J.; Kim, J.; Kim, J.-H.; Lee, J.; and Kim, S. 2024. Let 2D Diffusion Model Know 3D-Consistency for Robust Text-to-3D Generation. In ICLR
2024
-
[39]
Shi, Y.; Wang, P.; Ye, J.; Long, M.; Li, K.; and Yang, X. 2024. Mvdream: Multi-view diffusion for 3d generation. In ICLR
2024
-
[40]
Shonenkov, A.; Konstantinov, M.; Bakshandaeva, D.; Schuhmann, C.; Ivanova, K.; and Klokova, N. 2023. Deepfloyd. https://huggingface.co/DeepFloyd
2023
-
[41]
Song, Y.; and Ermon, S. 2019. Generative modeling by estimating gradients of the data distribution. Advances in neural information processing systems, 32
2019
-
[42]
Sun, J.; Fu, D.; Hu, Y.; Wang, S.; Rassin, R.; Juan, D.-C.; Alon, D.; Herrmann, C.; van Steenkiste, S.; Krishna, R.; et al. 2023. Dreamsync: Aligning text-to-image generation with image understanding feedback. In Synthetic Data for Computer Vision Workshop@ CVPR 2024
2023
-
[43]
Sun, Z.; Wu, T.; Zhang, P.; Zang, Y.; Dong, X.; Xiong, Y.; Lin, D.; and Wang, J. 2024. Bootstrap3D: Improving 3D Content Creation with Synthetic Data. arXiv preprint arXiv:2406.00093
2024 arXiv
-
[44]
Tang, J.; Ren, J.; Zhou, H.; Liu, Z.; and Zeng, G. 2024. Dreamgaussian: Generative gaussian splatting for efficient 3d content creation. In ICLR
2024
-
[45]
Touvron, H.; Lavril, T.; Izacard, G.; Martinet, X.; Lachaux, M.-A.; Lacroix, T.; Rozi \`e re, B.; Goyal, N.; Hambro, E.; Azhar, F.; et al. 2023. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971
2023 arXiv
-
[46]
Wang, Z.; Lu, C.; Wang, Y.; Bao, F.; Li, C.; Su, H.; and Zhu, J. 2024. Prolificdreamer: High-fidelity and diverse text-to-3d generation with variational score distillation. In NeurIPS
2024
-
[47]
Wu, S.; Lin, Y.; Zhang, F.; Zeng, Y.; Xu, J.; Torr, P.; Cao, X.; and Yao, Y. 2024 a . Direct3d: Scalable image-to-3d generation via 3d latent diffusion transformer. Advances in Neural Information Processing Systems, 37: 121859--121881
2024
-
[48]
Wu, T.; Yang, G.; Li, Z.; Zhang, K.; Liu, Z.; Guibas, L.; Lin, D.; and Wetzstein, G. 2024 b . Gpt-4v (ision) is a human-aligned evaluator for text-to-3d generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 22227--22238
2024
-
[49]
Xu, J.; Liu, X.; Wu, Y.; Tong, Y.; Li, Q.; Ding, M.; Tang, J.; and Dong, Y. 2023. Imagereward: Learning and evaluating human preferences for text-to-image generation. Advances in Neural Information Processing Systems
2023
-
[50]
Yan, J.; Gao, Y.; Yang, Q.; Wei, X.; Xie, X.; Wu, A.; and Zheng, W.-S. 2025. DreamView: Injecting View-specific Text Guidance into Text-to-3D Generation. In ECCV
2025
-
[51]
Yi, T.; Fang, J.; Wang, J.; Wu, G.; Xie, L.; Zhang, X.; Liu, W.; Tian, Q.; and Wang, X. 2024. GaussianDreamer: Fast Generation from Text to 3D Gaussians by Bridging 2D and 3D Diffusion Models. In CVPR
2024
-
[52]
Zhao, Z.; Lai, Z.; Lin, Q.; Zhao, Y.; Liu, H.; Yang, S.; Feng, Y.; Yang, M.; Zhang, S.; Yang, X.; et al. 2025. Hunyuan3d 2.0: Scaling diffusion models for high resolution textured 3d assets generation. arXiv preprint arXiv:2501.12202
2025 arXiv
-
[53]
Zhou, X.; Ran, X.; Xiong, Y.; He, J.; Lin, Z.; Wang, Y.; Sun, D.; and Yang, M.-H. 2024. Gala3d: Towards text-to-3d complex scene generation via layout-guided generative gaussian splatting. arXiv preprint arXiv:2402.07207
2024 arXiv
-
[54]
Zhu, J.; Zhuang, P.; and Koyejo, S. 2024. HiFA: High-fidelity Text-to-3D Generation with Advanced Diffusion Guidance. In ICLR
2024
Reviewed August 16, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.