Pith. sign in

REVIEW 3 major objections 5 minor 300 references

Multimodal models reason better about space when they draw and re-render their own intermediate SVG sketches.

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 · grok-4.5

2026-07-12 01:50 UTC pith:WJMEVWMN

load-bearing objection SVG as a revisable intermediate medium is a clean, useful idea with large gains; the main soft spot is RL on the same benchmark families, not a collapse of the claim. the 3 major comments →

arxiv 2607.03530 v1 pith:WJMEVWMN submitted 2026-07-03 cs.AI

MentalThink: Shaping Thoughts in Mental SVG World

Kangheng Lin , Jisheng Yin , Dingming Li , En Yu , Yana Wei , Han Zhou , Liang Zhao , Hongyu Zhou
show 6 more authors
Hongbo Peng Jianjian Sun Zheng Ge Xiangyu Zhang Daxin Jiang Jingyu Wang
This is my paper
classification cs.AI
keywords MentalThinkthink-with-SVGmultimodal reasoningscalable vector graphicsspatial reasoningmulti-turn RLmental imageryvisual-symbolic
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.

MentalThink argues that multimodal language models can do stronger spatial reasoning if they treat scalable vector graphics (SVG) code as an executable mental scratchpad. Instead of describing a scene only in words or sparse boxes, the model writes SVG, has it rendered into an image, looks at the result, and revises the sketch over multiple turns. A two-stage recipe first teaches the model to produce valid SVG (supervised fine-tuning) and then uses multi-turn reinforcement learning to reward iterative inspection and self-correction. On spatial benchmarks the method lifts a 7B backbone to 55.1% on VSIBench and 76.0% on MindCube, and qualitative traces show the model spontaneously taking new perspectives, fixing orientation errors, and composing multi-view scenes. The claim is that a deterministic, compositional visual language gives models a verifiable workspace that pure text lacks, so spatial hypotheses can be checked rather than merely narrated.

Core claim

Executable SVG code functions as a structured intermediate representation that lets multimodal models externalize spatial hypotheses, render them deterministically, and refine them multi-turn, producing large gains on geometry-heavy spatial benchmarks and eliciting dynamic perspective taking, visual reflection, and compositional scene construction that language-only or bounding-box intermediates do not match.

What carries the argument

The think-with-SVG pipeline: an SVG-augmented reasoning state that interleaves linguistic steps with optional SVG code, a deterministic renderer that turns code into an image observation, and a multi-turn RL loop (format reward plus answer reward) that rewards the model for generating, inspecting, and revising those mental sketches.

Load-bearing premise

That teacher-synthesized multi-turn SVG traces plus a simple format-and-answer reward are enough to produce genuine geometric self-correction rather than style-matching or benchmark-prompt memorization.

What would settle it

Train an identical backbone with matched multi-turn text or bounding-box trajectories (same teacher, same tasks, same RL budget) and check whether the SVG version still shows the reported 40-point MindCube and 24-point VSIBench gains and the same adaptive multi-turn revision patterns on held-out spatial problems.

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

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

3 major / 5 minor

Summary. MentalThink proposes a think-with-SVG pipeline that lets multimodal LLMs externalize spatial hypotheses as executable SVG code, render them deterministically, and condition subsequent reasoning on the rendered images. The method is realized by a two-stage recipe: SFT for syntactic alignment and SVG-mediated thought externalization, followed by multi-turn GRPO (Eq. 10) that rewards format validity and final-answer correctness. On a Qwen2.5-VL-7B backbone the approach reports large gains on spatial benchmarks (55.1% VSIBench, 76.0% MindCube-Tiny; Table 1), with ablations isolating SVG data, general spatial data, and RL (Table 2), a controlled paradigm comparison against text- and bbox-based intermediates (Table 3), and qualitative traces of perspective taking, visual refinement, and scene construction (Figure 3).

Significance. If the gains reflect a transferable geometric workspace rather than benchmark-specific pattern matching, the work supplies a clean, tool-free intermediate representation that is both linguistically native to LLMs and deterministically visualizable. The formalization of SVG-augmented states (Definitions 3.4–3.6), the closed render–inspect loop, and the controlled paradigm comparison (Table 3) are concrete contributions that other multimodal reasoning systems can build on. The adaptive System-1/System-2 statistics (Table 4) further suggest that the model can allocate visual computation according to task difficulty, which is a useful empirical finding for test-time scaling of spatial reasoning.

major comments (3)
  1. Section 4.1 and Appendix A.1 state that the RL corpus is drawn from the training splits of VSIBench and MindCube—the same two benchmarks that dominate the headline numbers in Table 1. The reward (Eq. 10) contains only format validity and final-answer match; there is no geometric-consistency, multi-view-consistency, or intermediate-SVG quality term. Consequently the large RL increments (especially MindCube 36%→76%) and the multi-turn revision patterns in Figure 3 remain compatible with the policy learning to emit teacher-style SVG sketches that co-occur with correct answers on these particular prompt distributions, rather than acquiring a general inspect-and-revise geometric capability. A held-out spatial suite never seen in RL (or an ablation that freezes the SVG generator after SFT and only optimizes answer selection) is needed to separate these explanations.
  2. Table 3’s paradigm comparison (Text / BBox / SVG) is the cleanest evidence that the intermediate representation itself matters, yet it is reported only for a single teacher-generated SFT setting and without error bars or multiple seeds. Because the subsequent RL stage re-uses the same benchmark families, it is unclear whether the SVG advantage survives once the policy is optimized under the format+answer reward. Reporting the same three-way comparison after RL (or at least variance across seeds) would make the central claim load-bearing rather than suggestive.
  3. The qualitative claim of “iterative inspection, revision, and refinement” (Abstract, §3.3, Figure 3) is supported only by selected traces. Table 4 shows that average turns rise only from 1.19 to 1.64 and that SVG is invoked on 63.7% of compositional items; there is no quantitative measure of how often an intermediate SVG is actually revised after rendering (as opposed to a single successful sketch followed by an answer). Without such a revision-rate or self-correction statistic, the multi-turn RL narrative remains under-supported relative to the performance claims.
minor comments (5)
  1. Table 1 reports MindCube* (MindCube-Tiny) without stating the exact subset size or selection criterion; a one-sentence clarification would aid reproducibility.
  2. The reward coefficients λ_fmt=0.25 / λ_ans=0.75 and T_max=5 are given without sensitivity analysis; a short appendix note on robustness would strengthen confidence in the free parameters.
  3. Figure 2’s example trajectory is helpful, but the rendered <image1–3> thumbnails are too small to verify that the SVG actually encodes the claimed egocentric axes; larger insets or an appendix figure would help.
  4. Several related-work citations (e.g., Visual Sketchpad, cognitive-map methods) are discussed at a high level; a short paragraph contrasting the executable, code-native nature of SVG against non-executable sketch or map tokens would sharpen the novelty claim.
  5. Typographical inconsistencies appear in model names (LLaV A vs LLaVA) and in the arXiv header date; a light copy-edit pass is warranted.

Circularity Check

0 steps flagged

Empirical SFT+RL pipeline on public benchmarks; only minor train-split reuse for RL prompts, not a by-construction tautology.

full rationale

MentalThink is an empirical methods paper: it defines a think-with-SVG state (Defs. 3.4–3.6), trains via SFT likelihood (Eq. 7) then GRPO with hybrid reward (Eq. 10: format validity + final-answer match), and reports measured accuracies on held-out splits of VSIBench, MindCube, etc. (Table 1). No first-principles derivation, uniqueness theorem, or fitted parameter is re-labeled as a “prediction.” The reward does not encode geometric self-consistency of intermediate SVGs, so the claim that RL “encourages iterative inspection” is interpretive rather than forced by construction; that is a causal-strength issue, not circularity. RL prompts are drawn from the training splits of the same two benchmarks that dominate the headline numbers (Sec. 4.1, A.1), which is a standard (if imperfect) practice and creates a mild contamination risk, but the evaluation remains against external baselines and official test splits. Data synthesis uses an external teacher (Gemini 3) and is filtered for renderability; this is ordinary distillation, not self-definitional. No load-bearing self-citation of an unverified uniqueness result appears. Hence only minor circularity risk (score 2); the central performance claims are externally falsifiable measurements, not tautologies.

Axiom & Free-Parameter Ledger

4 free parameters · 3 axioms · 2 invented entities

The central performance claim rests on standard ML training assumptions plus a handful of design choices (reward weights, horizon, teacher model) that are free parameters of the method rather than derived quantities. No new physical or mathematical entities are postulated beyond the engineering construct of the SVG environment.

free parameters (4)
  • format/answer reward weights (λ_fmt=0.25, λ_ans=0.75)
    Chosen by hand to balance syntactic validity against final correctness; not derived from first principles.
  • maximum reasoning horizon T_max=5
    Hard-coded cap on multi-turn refinement; affects both compute and the System-2 statistics reported.
  • SFT and RL learning rates / batch sizes
    5e-5 / 256 then 2e-6 / 64; standard but free hyper-parameters that control convergence.
  • volume and mix of synthetic SVG data (50k+50k+100k)
    Curated quantities and source mix that determine the inductive bias transferred from the teacher.
axioms (3)
  • domain assumption SVG code that is grammatically valid and renderable constitutes a complete and compositional representation of the spatial layouts needed for the target benchmarks.
    Stated in Definition 3.4 and used throughout; not proved, only motivated by SVG’s design.
  • domain assumption A deterministic SVG renderer can be inserted into the model’s context without breaking end-to-end differentiability of the policy or introducing non-stationary visual noise.
    Required by the interactive loop (Eqs. 8–9); assumed true for the experimental environment.
  • domain assumption Group Relative Policy Optimization with the hybrid reward yields stable multi-turn visual self-correction.
    Taken from prior RL literature and applied without additional theoretical guarantee.
invented entities (2)
  • think-with-SVG pipeline / SVG-augmented reasoning state no independent evidence
    purpose: Provides the intermediate visual-symbolic medium that the policy learns to generate, render, and revise.
    Defined in Definitions 3.5–3.6; the core engineering contribution of the paper.
  • MentalThink two-stage training recipe no independent evidence
    purpose: Instantiates the pipeline via SFT for syntax then multi-turn RL for iterative refinement.
    Described in Section 3.3; no external validation outside the reported experiments.

pith-pipeline@v1.1.0-grok45 · 26284 in / 2558 out tokens · 62629 ms · 2026-07-12T01:50:11.613697+00:00 · methodology

0 comments
read the original abstract

We introduce MentalThink, a visual-symbolic reasoning paradigm that equips Multimodal LLMs (MLLMs) with an executable mechanism for "mental" visualization. The core of MentalThink is a think-with-SVG pipeline, where the model learns to generate, render, and interpret scalable vector graphics (SVG) code as an intermediate visual representation for multi-turn reasoning. By creating structured vector sketches, the model can externalize spatial hypotheses, inspect them through deterministic rendering, and reason within a constrained geometric space, effectively mimicking the human process of mental imagery. We instantiate this paradigm through a two-stage training framework, combining Supervised Fine-Tuning (SFT) for SVG syntactic alignment with multi-turn Reinforcement Learning (RL) to encourage iterative inspection, revision, and refinement of intermediate visual hypotheses. Extensive evaluations demonstrate that MentalThink achieves superior performance on spatial understanding and reasoning benchmarks (e.g., 55.1% on VSIBench, 76.0% on MindCube), showing that executable vector graphics provide a verifiable visual workspace for dynamic perspective taking, visual reflection, and compositional scene construction.

Figures

Figures reproduced from arXiv: 2607.03530 by Daxin Jiang, Dingming Li, En Yu, Han Zhou, Hongbo Peng, Hongyu Zhou, Jianjian Sun, Jingyu Wang, Jisheng Yin, Kangheng Lin, Liang Zhao, Xiangyu Zhang, Yana Wei, Zheng Ge.

Figure 1
Figure 1. Figure 1: Conceptual comparison of different thinking paradigms. standing, transcending the surface-level pattern matching often observed in pure text generation. Second, being inher￾ently code-based, this representation aligns natively with the architecture of LLMs. Since models are extensively pretrained on both large-scale text corpora and markup or programming languages (e.g., HTML/XML), they can read￾ily reuse … view at source ↗
Figure 2
Figure 2. Figure 2: Think-with-SVG Pipeline. MentalThink trains MLLMs to natively use SVG code to render “mental images” during multi-turn thinking process with SVG environment, enabling more advanced and effective reasoning for complex problem-solving. The rendering pipeline R consists of three sequential trans￾formations: σ Parse −−−→ TDOM Layout −−−→ G = {g1, . . . , gm} Rasterize −−−−−→ Iσ (6) where TDOM is the document o… view at source ↗
Figure 3
Figure 3. Figure 3: Examples of Spatial Reasoning Patterns. MENTALTHINK demonstrates dynamic perspective taking, reflective visual refinement, and compositional scene construction. time compute into rigorous visual grounding when faced with intricate spatial logic. 5.2. Case Analysis: Emergent Spatial Reasoning via Think-with-SVG [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Additional case studies for the SVG reconstruction capability of MENTALTHINK. After listing bounding boxes, explicitly reason about spatial relations (left/right/front/behind/overlapping) based on the boxes. After <think>, output the final answer in **\\ boxed{}**. A.3. More Case Studies We present more case studies to illustrate the SVG recon￾struction capability ( [PITH_FULL_IMAGE:figures/full_fig_p015_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Additional case studies for think-with-SVG reasoning of MENTALTHINK (1/2). 16 [PITH_FULL_IMAGE:figures/full_fig_p016_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Additional case studies for think-with-SVG reasoning of MENTALTHINK (2/2). 17 [PITH_FULL_IMAGE:figures/full_fig_p017_6.png] 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

300 extracted references · 148 linked inside Pith

  1. [1]

    Scaling Learning Algorithms Towards

    Bengio, Yoshua and LeCun, Yann , booktitle =. Scaling Learning Algorithms Towards

  2. [2]

    and Osindero, Simon and Teh, Yee Whye , journal =

    Hinton, Geoffrey E. and Osindero, Simon and Teh, Yee Whye , journal =. A Fast Learning Algorithm for Deep Belief Nets , volume =

  3. [3]

    2016 , publisher=

    Deep learning , author=. 2016 , publisher=

    2016

  4. [4]

    International Conference on Machine Learning , pages=

    Zero-shot text-to-image generation , author=. International Conference on Machine Learning , pages=. 2021 , organization=

    2021

  5. [5]

    arXiv preprint arXiv:2206.10789 , volume=

    Scaling autoregressive models for content-rich text-to-image generation , author=. arXiv preprint arXiv:2206.10789 , volume=. 2022 , publisher=

    Pith/arXiv arXiv 2022

  6. [6]

    arXiv preprint arXiv:2110.04627 , year=

    Vector-quantized image modeling with improved vqgan , author=. arXiv preprint arXiv:2110.04627 , year=

    Pith/arXiv arXiv

  7. [7]

    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=

  8. [8]

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

    Autoregressive image generation using residual quantization , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  9. [9]

    Advances in neural information processing systems , volume=

    Neural discrete representation learning , author=. Advances in neural information processing systems , volume=

  10. [10]

    Advances in neural information processing systems , volume=

    Generating diverse high-fidelity images with vq-vae-2 , author=. Advances in neural information processing systems , volume=

  11. [11]

    arXiv preprint arXiv:1308.3432 , year=

    Estimating or propagating gradients through stochastic neurons for conditional computation , author=. arXiv preprint arXiv:1308.3432 , year=

    Pith/arXiv arXiv

  12. [12]

    arXiv preprint arXiv:1312.6114 , year=

    Auto-encoding variational bayes , author=. arXiv preprint arXiv:1312.6114 , year=

    Pith/arXiv arXiv

  13. [13]

    Advances in neural information processing systems , volume=

    Denoising diffusion probabilistic models , author=. Advances in neural information processing systems , volume=

  14. [14]

    Advances in neural information processing systems , volume=

    Generative modeling by estimating gradients of the data distribution , author=. Advances in neural information processing systems , volume=

  15. [15]

    arXiv preprint arXiv:2010.02502 , year=

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

    Pith/arXiv arXiv 2010

  16. [16]

    Advances in neural information processing systems , volume=

    Diffusion models beat gans on image synthesis , author=. Advances in neural information processing systems , volume=

  17. [17]

    Advances in Neural Information Processing Systems , volume=

    Dpm-solver: A fast ode solver for diffusion probabilistic model sampling in around 10 steps , author=. Advances in Neural Information Processing Systems , volume=

  18. [18]

    arXiv preprint arXiv:2207.12598 , year=

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

    Pith/arXiv arXiv

  19. [19]

    The Journal of Machine Learning Research , volume=

    Cascaded diffusion models for high fidelity image generation , author=. The Journal of Machine Learning Research , volume=. 2022 , publisher=

    2022

  20. [20]

    arXiv preprint arXiv:2204.06125 , volume=

    Hierarchical text-conditional image generation with clip latents , author=. arXiv preprint arXiv:2204.06125 , volume=

    Pith/arXiv arXiv

  21. [21]

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

    High-resolution image synthesis with latent diffusion models , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  22. [22]

    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=

  23. [23]

    Advances in Neural Information Processing Systems , volume=

    Photorealistic text-to-image diffusion models with deep language understanding , author=. Advances in Neural Information Processing Systems , volume=

  24. [24]

    arXiv preprint arXiv:2305.18295 , year=

    Raphael: Text-to-image generation via large mixture of diffusion paths , author=. arXiv preprint arXiv:2305.18295 , year=

    Pith/arXiv arXiv

  25. [25]

    arXiv preprint arXiv:2112.10741 , year=

    Glide: Towards photorealistic image generation and editing with text-guided diffusion models , author=. arXiv preprint arXiv:2112.10741 , year=

    Pith/arXiv arXiv

  26. [26]

    arXiv preprint arXiv:2307.01952 , year=

    Sdxl: Improving latent diffusion models for high-resolution image synthesis , author=. arXiv preprint arXiv:2307.01952 , year=

    Pith/arXiv arXiv

  27. [27]

    arXiv preprint arXiv:2310.00426 , year=

    PixArt: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis , author=. arXiv preprint arXiv:2310.00426 , year=

    Pith/arXiv arXiv

  28. [28]

    arXiv preprint arXiv:2312.04557 , year=

    GenTron: Delving Deep into Diffusion Transformers for Image and Video Generation , author=. arXiv preprint arXiv:2312.04557 , year=

    Pith/arXiv arXiv

  29. [29]

    2024 , eprint=

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

    2024

  30. [30]

    arXiv preprint arXiv:2303.01469 , year=

    Consistency models , author=. arXiv preprint arXiv:2303.01469 , year=

    Pith/arXiv arXiv

  31. [31]

    Computer Science

    Improving image generation with better captions , author=. Computer Science. https://cdn. openai. com/papers/dall-e-3. pdf , volume=

  32. [32]

    5: Three Insights towards Enhancing Aesthetic Quality in Text-to-Image Generation , author=

    Playground v2. 5: Three Insights towards Enhancing Aesthetic Quality in Text-to-Image Generation , author=. arXiv preprint arXiv:2402.17245 , year=

    Pith/arXiv arXiv

  33. [33]

    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=

  34. [34]

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

    Mage: Masked generative encoder to unify representation learning and image synthesis , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  35. [35]

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

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

  36. [36]

    Advances in Neural Information Processing Systems , volume=

    Movq: Modulating quantized vectors for high-fidelity image generation , author=. Advances in Neural Information Processing Systems , volume=

  37. [37]

    arXiv preprint arXiv:2309.15505 , year=

    Finite scalar quantization: Vq-vae made simple , author=. arXiv preprint arXiv:2309.15505 , year=

    Pith/arXiv arXiv

  38. [38]

    arXiv preprint arXiv:2310.05737 , year=

    Language Model Beats Diffusion--Tokenizer is Key to Visual Generation , author=. arXiv preprint arXiv:2310.05737 , year=

    Pith/arXiv arXiv

  39. [39]

    International Conference on Learning Representations , year=

    Phenaki: Variable length video generation from open domain textual descriptions , author=. International Conference on Learning Representations , year=

  40. [40]

    arXiv preprint arXiv:2301.00704 , year=

    Muse: Text-to-image generation via masked generative transformers , author=. arXiv preprint arXiv:2301.00704 , year=

    Pith/arXiv arXiv

  41. [41]

    arXiv preprint arXiv:2404.02905 , year=

    Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction , author=. arXiv preprint arXiv:2404.02905 , year=

    Pith/arXiv arXiv

  42. [42]

    Advances in neural information processing systems , volume=

    Generative adversarial nets , author=. Advances in neural information processing systems , volume=

  43. [43]

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

    Image-to-image translation with conditional adversarial networks , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  44. [44]

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

    A style-based generator architecture for generative adversarial networks , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  45. [45]

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

    Analyzing and improving the image quality of stylegan , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  46. [46]

    Advances in Neural Information Processing Systems , volume=

    Alias-free generative adversarial networks , author=. Advances in Neural Information Processing Systems , volume=

  47. [47]

    ACM SIGGRAPH 2022 conference proceedings , pages=

    Stylegan-xl: Scaling stylegan to large diverse datasets , author=. ACM SIGGRAPH 2022 conference proceedings , pages=

    2022

  48. [48]

    arXiv preprint arXiv:2301.09515 , year=

    Stylegan-t: Unlocking the power of gans for fast large-scale text-to-image synthesis , author=. arXiv preprint arXiv:2301.09515 , year=

    Pith/arXiv arXiv

  49. [49]

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

    Scaling up gans for text-to-image synthesis , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  50. [50]

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

    Unpaired image-to-image translation using cycle-consistent adversarial networks , author=. Proceedings of the IEEE international conference on computer vision , pages=

  51. [51]

    arXiv preprint arXiv:1809.11096 , year=

    Large scale GAN training for high fidelity natural image synthesis , author=. arXiv preprint arXiv:1809.11096 , year=

    Pith/arXiv arXiv

  52. [52]

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

    The unreasonable effectiveness of deep features as a perceptual metric , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  53. [53]

    Advances in neural information processing systems , volume=

    Gans trained by a two time-scale update rule converge to a local nash equilibrium , author=. Advances in neural information processing systems , volume=

  54. [54]

    Advances in neural information processing systems , volume=

    Improved techniques for training gans , author=. Advances in neural information processing systems , volume=

  55. [55]

    arXiv preprint arXiv:2103.03841 , year=

    Generating images with sparse representations , author=. arXiv preprint arXiv:2103.03841 , year=

    Pith/arXiv arXiv

  56. [56]

    Advances in neural information processing systems , volume=

    Improved precision and recall metric for assessing generative models , author=. Advances in neural information processing systems , volume=

  57. [57]

    Advances in neural information processing systems , volume=

    Attention is all you need , author=. Advances in neural information processing systems , volume=

  58. [58]

    arXiv preprint arXiv:1810.04805 , year=

    Bert: Pre-training of deep bidirectional transformers for language understanding , author=. arXiv preprint arXiv:1810.04805 , year=

    Pith/arXiv arXiv

  59. [59]

    The Journal of Machine Learning Research , volume=

    Exploring the limits of transfer learning with a unified text-to-text transformer , author=. The Journal of Machine Learning Research , volume=. 2020 , publisher=

    2020

  60. [60]

    article , year=

    Improving language understanding by generative pre-training , author=. article , year=

  61. [61]

    OpenAI blog , volume=

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

  62. [62]

    Advances in neural information processing systems , volume=

    Language models are few-shot learners , author=. Advances in neural information processing systems , volume=

  63. [63]

    Advances in Neural Information Processing Systems , volume=

    Training language models to follow instructions with human feedback , author=. Advances in Neural Information Processing Systems , volume=

  64. [64]

    arXiv preprint arXiv:2303.08774 , year=

    Gpt-4 technical report , author=. arXiv preprint arXiv:2303.08774 , year=

    Pith/arXiv arXiv

  65. [65]

    GPT-4V(ision) system card , author =

  66. [66]

    International conference on machine learning , pages=

    Generative pretraining from pixels , author=. International conference on machine learning , pages=. 2020 , organization=

    2020

  67. [67]

    2022 , howpublished =

    OpenAI , title =. 2022 , howpublished =

    2022

  68. [68]

    2023 , howpublished =

    Google , title =. 2023 , howpublished =

    2023

  69. [69]

    2023 , howpublished =

    Anthropic , title =. 2023 , howpublished =

    2023

  70. [70]

    arXiv preprint arXiv:2302.13971 , year=

    Llama: Open and efficient foundation language models , author=. arXiv preprint arXiv:2302.13971 , year=

    Pith/arXiv arXiv

  71. [71]

    arXiv preprint arXiv:2307.09288 , year=

    Llama 2: Open foundation and fine-tuned chat models , author=. arXiv preprint arXiv:2307.09288 , year=

    Pith/arXiv arXiv

  72. [72]

    arXiv preprint arXiv:2203.15556 , year=

    Training compute-optimal large language models , author=. arXiv preprint arXiv:2203.15556 , year=

    Pith/arXiv arXiv

  73. [73]

    Journal of Machine Learning Research , volume=

    Palm: Scaling language modeling with pathways , author=. Journal of Machine Learning Research , volume=

  74. [74]

    arXiv preprint arXiv:2305.10403 , year=

    Palm 2 technical report , author=. arXiv preprint arXiv:2305.10403 , year=

    Pith/arXiv arXiv

  75. [75]

    arXiv preprint arXiv:2211.05100 , year=

    Bloom: A 176b-parameter open-access multilingual language model , author=. arXiv preprint arXiv:2211.05100 , year=

    Pith/arXiv arXiv

  76. [76]

    arXiv preprint arXiv:2205.01068 , year=

    Opt: Open pre-trained transformer language models , author=. arXiv preprint arXiv:2205.01068 , year=

    Pith/arXiv arXiv

  77. [77]

    arXiv preprint arXiv:2309.16609 , year=

    Qwen technical report , author=. arXiv preprint arXiv:2309.16609 , year=

    Pith/arXiv arXiv

  78. [78]

    arXiv preprint arXiv:2309.10305 , year=

    Baichuan 2: Open large-scale language models , author=. arXiv preprint arXiv:2309.10305 , year=

    Pith/arXiv arXiv

  79. [79]

    2023-01-06)[2023-09-27]

    Internlm: A multilingual language model with progressively enhanced capabilities , author=. 2023-01-06)[2023-09-27]. https://github. com/InternLM/InternLM , year=

    2023

  80. [80]

    arXiv preprint arXiv:2401.02954 , year=

    Deepseek llm: Scaling open-source language models with longtermism , author=. arXiv preprint arXiv:2401.02954 , year=

    Pith/arXiv arXiv

Showing first 80 references.