Pith. sign in

REVIEW 5 major objections 5 minor 41 references

NeuroGen: Neural Network Parameter Generation via Large Language Models

T0 review · 5 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read An LLM can generate functional neural network parameters directly from task and data descriptions, with no gradient-based optimization of the target network.

desk verdict A genuinely new proof-of-concept for LLM-generated weights, but the low-data advantage likely leaks full-dataset information and the 'no gradient' framing is overstated. read the letter →

arxiv 2505.12470 v2 pith:UDOGGNPV submitted 2025-05-18 cs.AI

classification cs.AI
keywords neuralnetworkparametergenerationlargelanguagemodelsreferenceknowledgeinjectioncontext-enhancedinstructiontuningprompt-drivenweightnon-autoregressivesynthesisclassificationbenchmarks
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

NeuroGen tries to establish a new route to neural network weights: instead of fitting parameters to data with backpropagation, a large language model produces the whole parameter vector in one non-autoregressive pass, conditioned on a description of the task, the data, and the target architecture. The method is two-stage: first the LLM is aligned to a distribution of conventionally trained checkpoints through a similarity loss, then it is instruction-tuned so that the weights it emits minimize the task loss on a sample of real training data. The experiments show generated classifiers reaching 97.71% accuracy on MNIST and 76.03% on SST-2, close to or better than classically trained counterparts on some benchmarks. The paper interprets this as evidence that LLMs can internalize a latent mapping from training context to weights, opening the possibility of prompt-driven, data-efficient model generation.

What carries the argument

The load-bearing object is a learnable special-token matrix $P\in\mathbb{R}^{d_1\times d_2}$ that is concatenated with the instruction embedding, passes through the LLM, and is mapped by a projection MLP to a flat parameter vector, which is then sliced into layer-wise weights of the target network $f$. Two objectives drive training: Stage 1 aligns generated weights $w_g$ to reference checkpoints $w_i$ via $L_1(P,\phi,\theta)=\mathbb{E}_{w_i\sim W}[\operatorname{sim}(w_i,w_g)]$, for example negative mean squared error; Stage 2 minimizes $L_2(P,\phi,\theta)=\sum_{(x_j,y_j)\in D_{\mathrm{sub}}^t} \mathcal{L}_{\mathrm{task}}(f(x_j\mid w_g^t),y_j)$. Gradients update the special token $P$, the LoRA parameters $\phi$, and the projection parameters $\theta$; the target network's own weights $w_g^t$ are outputs, not optimized variables.

What would settle it

Re-run the Figure 3 generalization experiment with Stage-1 checkpoints trained only on datasets disjoint from SVHN and CIFAR-10 (or on synthetic images), then apply Stage 2 on 10,000 samples of SVHN and CIFAR-10; if NeuroGen still outperforms a classical model trained on those 10,000 samples the data-efficiency claim survives, and if accuracy collapses the limited-data advantage came from prior exposure to full-data checkpoints of the same datasets.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that an LLM can be turned into a conditional generator of usable neural network parameters. The target network's weights are never optimized directly; instead, the LLM's learnable special token, LoRA adapters, and a projection MLP are optimized so that the emitted weight vector performs well on the task. Stage 1 supplies the needed prior by regressing generated weights toward checkpoints obtained by ordinary gradient-based training, and Stage 2 refines that prior by backpropagating the target task's cross-entropy through the instantiated generated network into the LLM's parameters. The reported accuracies on standard classification benchmarks are the evidence that the generated parameters are functionally effective, not merely distributionally plausible.

Load-bearing premise

The claim that NeuroGen is data-efficient assumes the Stage-1 checkpoints were not trained on the same full datasets used to judge limited-data generalization, because the paper does not report what data those larger-CNN checkpoints were trained on.

Editorial extensions

If this is right

  • A single LLM can act as a reusable weight generator: Stage-1 parameter-reference knowledge transfers across tasks, and only Stage 2 has to be repeated for a new task or dataset.
  • In small-data settings, generated weights can beat weights obtained by classical training on the same sample, which would make LLM priors a practical substitute for training data.
  • Deployment could become prompt-driven: rather than shipping fitted weights, a system could ship the architecture description and generate task-specific weights at inference time on the device.
  • The non-autoregressive, full-weight generation strategy limits target-network size, so scaling to large models would require generating only adapters, prompts, or low-rank updates rather than the full parameter vector.

Reading between the lines

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

  • A natural extension not tested here: hold out entire classes or datasets from Stage-1 pretraining and ask whether generated weights on those held-out classes still beat classical training; this would separate memorization of a weight distribution from compositional generation.
  • Another untested design choice: replacing the Stage-1 MSE alignment with a functional loss (such as agreement of logits on unlabeled data) might improve generated weights because it optimizes behavior rather than raw parameter coordinates.
  • If the mapping from task/data descriptions to weights is genuinely learnable, then inspecting the generated weights becomes a form of interpretability—the LLM's output is its answer to what a classifier for the described task should look like.
  • Since weights are produced in one forward pass, changing the instruction or data subset changes the weights, suggesting a testable route to model editing by prompt rather than by retraining.
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

5 major / 5 minor

Summary. The paper proposes NeuroGen, a two-stage framework in which an LLM is fine-tuned to generate the parameters of a lightweight target neural network directly, given a task description and training data. In Stage 1, the LLM is aligned with reference checkpoints of the target architecture through learnable special tokens, LoRA, and a projection MLP, using a parameter-space similarity loss. In Stage 2, the model is instruction-tuned with a task loss evaluated through the generated weights on a subset of training data. Experiments cover image classification (MNIST, SVHN, CIFAR-10) with CNN and LeNet targets and text classification (SST-2, SNLI, AG News) with MLP and RNN targets. The main empirical claim is that NeuroGen produces usable neural network parameters, competitive with or sometimes better than standard gradient-based training, and that after Stage-1 pretraining on a larger architecture, Stage-2-only tuning can generate smaller models that outperform classical training in low-data regimes.

Significance. If the central claim holds, the paper opens a distinct research direction: treating neural network weight acquisition as a conditional generation problem solved by LLMs. The framework is clearly described, explicitly limited to small classification models, and accompanied by an honest limitation section. The strengths include a transparent two-stage training scheme, the use of non-autoregressive parallel parameter generation, and a direct comparison against classical training on standard benchmarks. However, the empirical support is currently thin: the headline results are single runs with no variance, no code is released, the connection to the diffusion-based weight generation literature is asserted rather than experimentally contrasted, and the low-data generalization experiment in Section 4.4 is vulnerable to information leakage because the Stage-1 checkpoint source is unspecified. The claimed 'without gradient-based optimization' framing is also stronger than what Equation (2) actually implements, since the generator parameters are optimized by gradient descent.

major comments (5)
  1. [§4.4, Figure 3] The limited-data comparison in Section 4.4 is load-bearing for the paper's data-efficiency claim, but the paper never states what data was used to obtain the Stage-1 reference checkpoints of the larger CNN. If those checkpoints were trained on the full SVHN or CIFAR-10 training sets, then the LLM was pre-exposed to label-dependent information from the full data before Stage 2 was run on only 10,000 samples, making the comparison against a classical model trained only on 10,000 samples unfair. The authors must specify the training data for the Stage-1 larger-CNN checkpoints, and ideally rerun the comparison with Stage-1 pretraining also restricted to the 10,000-sample subset, to establish that the observed advantage comes from generative generalization rather than from memorization of the full dataset.
  2. [§2.1, Eq. (2), Contribution 1] The contribution 'without gradient-based optimization' is contradicted by Equation (2), where the parameters {P, phi, theta} are explicitly updated by minimizing the task loss L2 via gradient-based optimization, and by Algorithm 1's update steps. What the paper actually demonstrates is that the target weights w_g are not directly trained, while the generator that produces them is trained with gradients. The claim should be rephrased as 'without direct gradient-based optimization of the target network parameters' to be accurate, and the discussion in Sections 1 and 5 should be adjusted accordingly.
  3. [Table 1, §4.2] All six accuracy numbers in Table 1 are reported as single values with no error bars, no number of seeds, and no statistical significance tests. Because the differences between NeuroGen and the classical baseline are small in several cases (e.g., SST-2 RNN: 76.03 vs. 77.63, AG News RNN: 85.14 vs. 84.72), the central empirical claim that LLM-generated parameters are 'usable' and comparable to classical training is not yet established with confidence. The authors should report means and standard deviations over at least three to five seeds, and ideally provide per-seed results.
  4. [§3.3–§3.4, §4.1] Several details needed to reproduce NeuroGen are missing: the size of the Stage-2 data subset |D_sub|, the LoRA rank, the projection MLP architecture and hidden dimension, the number of special tokens d1, the exact learning rate schedule for each stage, and the optimizer details beyond 'SGD optimizer'. Additionally, footnote 2 states that code will be released only after acceptance, which further impedes verification. These omissions are significant because the paper's contribution is described as an 'easy-to-implement' framework, and the reported results cannot be checked without them.
  5. [§2.3, §4.2] The paper dismisses hypernetwork and diffusion-based weight-generation baselines as 'not directly applicable or comparable' in Section 2.3, but does not provide any experimental comparison with them. Since Neural Network Diffusion, Conditional LoRA Parameter Generation, and hypernetworks are the most closely related prior methods, the significance of NeuroGen relative to those approaches is unclear. At minimum, the authors should include a small-scale comparison on one image and one text task, or provide a concrete argument with evidence for why such baselines cannot be adapted to the instruction-conditioned setting.
minor comments (5)
  1. [§4.1] The Implementation Details paragraph contains a leftover reference to 'the proposed FedType', which is not defined in this paper and appears to be a remnant of another project; it should be replaced with 'NeuroGen'.
  2. [Throughout] The manuscript contains numerous typos and misspellings, including 'approachaes', 'paramaters', 'nerual', 'mechnism', 'priliminary', and 'netwrok'. A thorough proofreading pass is needed.
  3. [Figure 2] The figure caption refers to 'Phase 1 and Phase' without specifying the second phase; it should read 'Phase 1 and Phase 2'.
  4. [Algorithm 1, Eq. (1)] Equation (1) defines the Stage-1 objective with a generic similarity function sim(·,·), while Algorithm 1 uses L_mse; these should be made consistent, and the exact similarity function used in the experiments should be stated.
  5. [Table 1] The table layout is confusing: the header 'NN Image NN Text' merges two separate sub-tables, and the row labels 'Classical CNN' and 'NeuroGen' are repeated for different architectures. Separating the image and text results into two tables or using clearer multi-level headers would improve readability.

Circularity Check

1 steps flagged · score 3.0 of 10

Low-data generalization claim in §4.4 inherits full-data Stage-1 checkpoints, making that comparison partially circular; main results are still independent.

  1. fitted input called prediction [Section 4.4 / Figure 3, with Section 3.3.1 and Section 4.5]
    "we construct a dataset of neural network checkpoints obtained through standard gradient descent training ... obtained by training on the full dataset D with a distinct random seed. ... we generate parameters for a smaller CNN using an LLM that was pre-trained to generate parameters for a larger CNN. Specifically, we apply only Stage 2 (context-enhanced tuning) to generate the smaller model. ..."

    Figure 3's limited-data result is offered as evidence that NeuroGen generates a competitive smaller CNN from only 10,000 samples. But by the paper's own Stage-1 protocol, the checkpoints used to pretrain the LLM are trained on 'the full dataset D'; for the SVHN/CIFAR-10 experiments in Figure 3, D is the full SVHN/CIFAR-10 training set. The LLM has therefore been aligned (Eq. 1) to parameters derived from all full-data labels before Stage 2 sees the 10k subset, so the reported low-data advantage is inherited from full-data fitting rather than predicted from the 10k samples alone. The paper never states the pretraining data for the larger CNN, making the comparison to a classical model trained only on 10k samples non-identifying.

full rationale

Aside from the low-data generalization claim in §4.4, I find no circularity. There are no load-bearing self-citations: references [7,8,9] are external prior work, not the authors' own theorems. Table 1 reports held-out test accuracy of models whose weights are produced by the LLM; no reported number is constructed from a fitted constant by definition. The 'without gradient-based optimization' contribution is qualified in §3.4.2 and Algorithm 1: the target weights w_t_g are not directly trained, but {P, phi, theta} are optimized via L2; this is a framing/novelty weakness, not a circular reduction, because the generated weights are not defined as the minimizer of the test loss. The only partially circular component is the §4.4 data-efficiency comparison, where the LLM used for the limited-data experiment was Stage-1 pretrained on checkpoints that, per §3.3.1, are trained on the full dataset D, and the paper does not disclose what data the larger-CNN pretraining used. Because the central Table 1 results are self-contained external measurements, the overall circularity is partial and localized.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The method depends on several unreported or under-specified numbers (N, subset size, LoRA rank, projection width) and on task-specific reference checkpoints in Stage 1. No new physical or scientific entities are introduced; the learnable special token P is a parameterized input, not a new entity.

free parameters (6)
  • Number of reference checkpoints N
    Stage 1 samples N checkpoints per dataset to define the reference distribution W, but N is never specified and the alignment objective L1 averages over it.
  • Stage-2 data subset size |D_sub|
    Section 3.4.1 says a small randomly sampled subset is used, but the size is never given; Section 4.2 says both methods use the full training set, which is inconsistent.
  • Training epochs in Stage 1 and Stage 2 = 30 and 20
    Chosen without ablation or sensitivity analysis; these numbers directly affect convergence and final accuracy.
  • Learning rate schedule = 1e-3, halved every 10 epochs
    A fixed SGD schedule chosen by hand; no sensitivity analysis is reported.
  • Soft-clipping hyperparameter alpha
    Section 4.3 introduces alpha to normalize parameters when Stage 1 is removed; the value or range is not stated.
  • LoRA rank and projection MLP dimensions
    Not reported; these determine the capacity that maps LLM hidden states to target weights and are central to generation quality.
assumptions (4)
  • domain assumption Neural network parameters encode recoverable information about training data and tasks, based on memorization and membership-inference studies.
    The entire premise that checkpoints can be aligned with data and task context via LLM generation rests on weight-data correlations cited in Section 1.
  • domain assumption A fixed small target architecture can be represented by a flat parameter vector that fits inside the LLM's hidden-state budget d1*d2 >= |w|.
    Section 3.3.1 assumes a linear projection can faithfully carry continuous weight values; no evidence is given that the vocabulary or hidden-state geometry of an LLM preserves this information.
  • ad hoc to paper Stage 1 reference checkpoints are trained on the same task and dataset as Stage 2 and are representative of the target distribution W.
    Algorithm 1 obtains reference weights from dataset D_t for the same task t; the method is not evaluated for generating parameters for unseen datasets or tasks.
  • domain assumption Checkpoints trained with different random seeds on the same data are a sufficient training signal for weight-space understanding.
    Stage 1 uses only these checkpoints and does not inject diversity in architectures, optimizers, or training schedules.

how reviews work

0 comments
Cite this review

Pith. "Pith review of NeuroGen: Neural Network Parameter Generation via Large Language Models." pith.science (2026). https://pith.science/paper/UDOGGNPV

@misc{pith2026250512470,
  author       = {Pith},
  title        = {Pith review of: NeuroGen: Neural Network Parameter Generation via Large Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UDOGGNPV}},
  note         = {Machine review of arXiv:2505.12470}
}
read the original abstract

Acquiring the parameters of neural networks (NNs) has been one of the most important problems in machine learning since the inception of NNs. Traditional approaches, such as backpropagation and forward-only optimization, acquire parameters via iterative data fitting to gradually optimize them. This paper aims to explore the feasibility of a new direction: acquiring NN parameters via large language model generation. We propose NeuroGen, a generalized and easy-to-implement two-stage approach for NN parameter generation conditioned on descriptions of the data, task, and network architecture. Stage one is Parameter Reference Knowledge Injection, where LLMs are pretrained on NN checkpoints to build foundational understanding of parameter space, whereas stage two is Context-Enhanced Instruction Tuning, enabling LLMs to adapt to specific tasks through enriched, task-aware prompts. Experimental results demonstrate that NeuroGen effectively generates usable NN parameters. Our findings highlight the feasibility of LLM-based NN parameter generation and suggest a promising new paradigm where LLMs and lightweight NNs can coexist synergistically

Figures

Figures reproduced from arXiv: 2505.12470 by the authors.

Figure 1
Figure 1. Framework of NeuroGen. robotics [22]. They have been successfully applied to various tasks such as question answering [23], reasoning [24], and data generation [25]. In particular, substantial research has been devoted to LLM-driven image generation [26], text generation [27], and code generation [28]. Multimodal LLMs combine the advanced capabilities of LLMs with additional modalities such as images, audio, and vid… view at source ↗
Figure 2
Figure 2. Ablation study on MNIST with CNN generation and AG News with MLP generation. [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Model generation study on SVHN and CIFAR-10 datasets using CNN generation. Training [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 16 canonical work pages

  1. [1]

    Learning representations by back-propagating errors.nature, 323(6088):533–536, 1986

    David E Rumelhart, Geoffrey E Hinton, and Ronald J Williams. Learning representations by back-propagating errors.nature, 323(6088):533–536, 1986

  2. [2]

    Imagenet classification with deep convolutional neural networks.Communications of the ACM, 60(6):84–90, 2017

    Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. Imagenet classification with deep convolutional neural networks.Communications of the ACM, 60(6):84–90, 2017

  3. [3]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 770–778, 2016

  4. [4]

    Long short-term memory.Supervised sequence labelling with recurrent neural networks, pages 37–45, 2012

    Alex Graves and Alex Graves. Long short-term memory.Supervised sequence labelling with recurrent neural networks, pages 37–45, 2012

  5. [5]

    Attention is all you need.Advances in neural information processing systems, 30, 2017

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need.Advances in neural information processing systems, 30, 2017

  6. [6]

    Bert: Pre-training of deep bidirectional transformers for language understanding

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep bidirectional transformers for language understanding. InProceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies, volume 1 (long and short papers), pages 4171–4186, 2019

  7. [7]

    Neural network diffusion.arXiv preprint arXiv:2402.13144, 2024

    Kai Wang, Dongwen Tang, Boya Zeng, Yida Yin, Zhaopan Xu, Yukun Zhou, Zelin Zang, Trevor Darrell, Zhuang Liu, and Yang You. Neural network diffusion.arXiv preprint arXiv:2402.13144, 2024

  8. [8]

    Conditional lora parameter generation.arXiv preprint arXiv:2408.01415, 2024

    Xiaolong Jin, Kai Wang, Dongwen Tang, Wangbo Zhao, Yukun Zhou, Junshu Tang, and Yang You. Conditional lora parameter generation.arXiv preprint arXiv:2408.01415, 2024

Show all 41 references
  1. [9]

    Learning to learn with generative models of neural network checkpoints.arXiv preprint arXiv:2209.12892, 2022

    William Peebles, Ilija Radosavovic, Tim Brooks, Alexei A Efros, and Jitendra Malik. Learning to learn with generative models of neural network checkpoints.arXiv preprint arXiv:2209.12892, 2022

  2. [10]

    Denoising diffusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020

  3. [11]

    Diffusion models beat gans on image synthesis

    Prafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis. Advances in neural information processing systems, 34:8780–8794, 2021

  4. [12]

    The secret sharer: Evaluating and testing unintended memorization in neural networks

    Nicholas Carlini, Chang Liu, Úlfar Erlingsson, Jernej Kos, and Dawn Song. The secret sharer: Evaluating and testing unintended memorization in neural networks. In28th USENIX security symposium (USENIX security 19), pages 267–284, 2019

  5. [13]

    Membership inference attacks against machine learning models

    Reza Shokri, Marco Stronati, Congzheng Song, and Vitaly Shmatikov. Membership inference attacks against machine learning models. In2017 IEEE symposium on security and privacy (SP), pages 3–18. IEEE, 2017

  6. [14]

    Auditing data provenance in text-generation models

    Congzheng Song and Vitaly Shmatikov. Auditing data provenance in text-generation models. InProceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pages 196–206, 2019

  7. [15]

    Language models are few-shot learners.Advances in neural information processing systems, 33:1877–1901, 2020

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learners.Advances in neural information processing systems, 33:1877–1901, 2020

  8. [16]

    Training language models to follow instructions with human feedback.Advances in neural information processing systems, 35:27730–27744, 2022

    Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. Training language models to follow instructions with human feedback.Advances in neural information processing systems, 35:27730...

  9. [17]

    Hypernetworks.arXiv preprint arXiv:1609.09106, 2016

    David Ha, Andrew Dai, and Quoc V Le. Hypernetworks.arXiv preprint arXiv:1609.09106, 2016

  10. [18]

    Model-agnostic meta-learning for fast adap- tation of deep networks

    Chelsea Finn, Pieter Abbeel, and Sergey Levine. Model-agnostic meta-learning for fast adap- tation of deep networks. InInternational conference on machine learning, pages 1126–1135. PMLR, 2017. 10

  11. [19]

    Recent advances in natural language processing via large pre-trained language models: A survey.ACM Computing Surveys, 56(2):1– 40, 2023

    Bonan Min, Hayley Ross, Elior Sulem, Amir Pouran Ben Veyseh, Thien Huu Nguyen, Oscar Sainz, Eneko Agirre, Ilana Heintz, and Dan Roth. Recent advances in natural language processing via large pre-trained language models: A survey.ACM Computing Surveys, 56(2):1– 40, 2023

  12. [20]

    Visionllm: Large language model is also an open-ended decoder for vision-centric tasks.Advances in Neural Information Processing Systems, 36:61501–61513, 2023

    Wenhai Wang, Zhe Chen, Xiaokang Chen, Jiannan Wu, Xizhou Zhu, Gang Zeng, Ping Luo, Tong Lu, Jie Zhou, Yu Qiao, et al. Visionllm: Large language model is also an open-ended decoder for vision-centric tasks.Advances in Neural Information Processing Systems, 36:61501–61513, 2023

  13. [21]

    A large language model for electronic health records.NPJ digital medicine, 5(1):194, 2022

    Xi Yang, Aokun Chen, Nima PourNejatian, Hoo Chang Shin, Kaleb E Smith, Christopher Parisien, Colin Compas, Cheryl Martin, Anthony B Costa, Mona G Flores, et al. A large language model for electronic health records.NPJ digital medicine, 5(1):194, 2022

  14. [22]

    Large language models for robotics: A survey.arXiv preprint arXiv:2311.07226, 2023

    Fanlong Zeng, Wensheng Gan, Yongheng Wang, Ning Liu, and Philip S Yu. Large language models for robotics: A survey.arXiv preprint arXiv:2311.07226, 2023

  15. [23]

    Questioning the survey responses of large language models.Advances in Neural Information Processing Systems, 37:45850–45878, 2024

    Ricardo Dominguez-Olmedo, Moritz Hardt, and Celestine Mendler-Dünner. Questioning the survey responses of large language models.Advances in Neural Information Processing Systems, 37:45850–45878, 2024

  16. [24]

    Reasoning with large language models, a survey.arXiv preprint arXiv:2407.11511, 2024

    Aske Plaat, Annie Wong, Suzan Verberne, Joost Broekens, Niki van Stein, and Thomas Back. Reasoning with large language models, a survey.arXiv preprint arXiv:2407.11511, 2024

  17. [25]

    A survey on data synthesis and augmentation for large language models.arXiv preprint arXiv:2410.12896, 2024

    Ke Wang, Jiahui Zhu, Minjie Ren, Zeming Liu, Shiwei Li, Zongye Zhang, Chenkai Zhang, Xiaoyu Wu, Qiqi Zhan, Qingjie Liu, et al. A survey on data synthesis and augmentation for large language models.arXiv preprint arXiv:2410.12896, 2024

  18. [26]

    Renaissance: A survey into ai text-to-image generation in the era of large model.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024

    Fengxiang Bie, Yibo Yang, Zhongzhu Zhou, Adam Ghanem, Minjia Zhang, Zhewei Yao, Xiaoxia Wu, Connor Holmes, Pareesa Golnari, David A Clifton, et al. Renaissance: A survey into ai text-to-image generation in the era of large model.IEEE Transactions on Pattern Analysis and Machin...

  19. [27]

    Pre-trained language models for text generation: A survey.ACM Computing Surveys, 56(9):1–39, 2024

    Junyi Li, Tianyi Tang, Wayne Xin Zhao, Jian-Yun Nie, and Ji-Rong Wen. Pre-trained language models for text generation: A survey.ACM Computing Surveys, 56(9):1–39, 2024

  20. [28]

    A survey on large language models for code generation.arXiv preprint arXiv:2406.00515, 2024

    Juyong Jiang, Fan Wang, Jiasi Shen, Sungju Kim, and Sunghun Kim. A survey on large language models for code generation.arXiv preprint arXiv:2406.00515, 2024

  21. [29]

    Multimodal machine learn- ing: A survey and taxonomy.IEEE Trans

    Tadas Baltrusaitis, Chaitanya Ahuja, and Louis-Philippe Morency. Multimodal machine learn- ing: A survey and taxonomy.IEEE Trans. Pattern Anal. Mach. Intell., 41(2):423–443, 2019

  22. [30]

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

    Junnan Li, Dongxu Li, Caiming Xiong, and Steven Hoi. Blip: Bootstrapping language-image pre-training for unified vision-language understanding and generation. InICML, 2022

  23. [31]

    Flamingo: a visual language model for few-shot learning

    Jean-Baptiste Alayrac, Jeff Donahue, Pauline Luc, Antoine Miech, Iain Barr, Yana Hasson, Karel Lenc, Arthur Mensch, Katie Millicah, Malcolm Reynolds, Roman Ring, Eliza Ruther- ford, Serkan Cabi, Tengda Han, Zhitao Gong, Sina Samangooei, Marianne Monteiro, Jacob Menick, Sebasti...

  24. [32]

    Visual instruction tuning

    Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. Visual instruction tuning. In Proceedings of the 37th International Conference on Neural Information Processing Systems, NIPS ’23, 2023

  25. [33]

    The mnist database of handwritten digit images for machine learning research.IEEE Signal Processing Magazine, 2012

    Li Deng. The mnist database of handwritten digit images for machine learning research.IEEE Signal Processing Magazine, 2012

  26. [34]

    Yuval Netzer, Tao Wang, Adam Coates, Alessandro Bissacco, Bo Wu, and Andrew Y . Ng. Reading digits in natural images with unsupervised feature learning. InNIPS Workshop on Deep Learning and Unsupervised Feature Learning, 2011

  27. [35]

    Learning multiple layers of features from tiny images

    Alex Krizhevsky and Hinton. Learning multiple layers of features from tiny images. http: //www.cs.toronto.edu/~kriz/learning-features-2009-TR.pdf, 2009

  28. [36]

    Manning, Andrew Y

    Richard Socher, Alex Perelygin, Jean Wu, Jason Chuang, Christopher D. Manning, Andrew Y . Ng, and Christopher Potts. Recursive deep models for semantic compositionality over a sentiment treebank. InEMNLP, 2013. 11

  29. [37]

    Bowman, Gabor Angeli, Christopher Potts, and Christopher D

    Samuel R. Bowman, Gabor Angeli, Christopher Potts, and Christopher D. Manning. A large annotated corpus for learning natural language inference. InEMNLP, pages 632–642, 2015

  30. [38]

    Character-level convolutional networks for text classification

    Xiang Zhang, Junbo Jake Zhao, and Yann LeCun. Character-level convolutional networks for text classification. InNeurIPS, 2015

  31. [39]

    Qwen2-vl: Enhancing vision-language model’s perception of the world at any resolution.arXiv preprint arXiv:2409.12191, 2024

    Peng Wang, Shuai Bai, Sinan Tan, Shijie Wang, Zhihao Fan, Jinze Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, et al. Qwen2-vl: Enhancing vision-language model’s perception of the world at any resolution.arXiv preprint arXiv:2409.12191, 2024

  32. [40]

    The llama 3 herd of models.arXiv preprint arXiv:2407.21783, 2024

    Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, et al. The llama 3 herd of models.arXiv preprint arXiv:2407.21783, 2024

  33. [41]

    Lecun, L

    Y . Lecun, L. Bottou, Y . Bengio, and P. Haffner. Gradient-based learning applied to document recognition.Proceedings of the IEEE, 86(11):2278–2324, 1998. 12

Pith tools

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