LVPruning: An Effective yet Simple Language-Guided Vision Token Pruning Approach for Multi-modal Large Language Models

Chenghua Lin; Hao Li; Jingyuan Sun; Riza Batista-Navarro; Yanze Xin; Yizheng Sun

REVIEW 4 major objections 5 minor 31 references

LVPruning: An Effective yet Simple Language-Guided Vision Token Pruning Approach for Multi-modal Large Language Models

T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A lightweight language-guided pruning module can remove up to 90% of vision tokens from a multimodal LLM while keeping average benchmark loss near 0.45%.

desk verdict A simple, plausible method for language-guided vision token pruning in MLLMs, but the headline accuracy and FLOPs claims are under-supported and the paper skips the obvious MLLM pruning baselines. read the letter →

arxiv 2501.13652 v2 pith:Y3MPM5MQ submitted 2025-01-23 cs.CL

Yizheng Sun , Yanze Xin , Hao Li , Jingyuan Sun , Chenghua Lin , Riza Batista-Navarro This is my paper · ORCID

classification cs.CL

keywords visiontokenpruningmultimodallargelanguagemodelsefficientinferencecross-attentionLLaVA-1.5sparsificationinstructiontuningGumbel-Softmax

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

This paper introduces LVPruning, a method for cutting the inference cost of multimodal large language models by pruning vision tokens early and progressively. The central claim is that language tokens—the prompt and partially generated text—are sufficient guides for deciding which image patches matter, so that up to 90% of vision tokens can be dropped by the middle of LLaVA-1.5 without retraining the base model. The reported result is a 62.1% reduction in inference TFLOPs (8.38 to 3.18) with an average performance loss of roughly 0.45% across nine benchmarks. This matters because it suggests that a large share of visual computation in MLLMs is redundant for the task, and that efficiency gains do not have to come at the cost of accuracy or from building a new architecture.

What carries the argument

The load-bearing component is the cross-attention decision module: a lightweight transformer block inserted at selected LLM layers where vision tokens act as queries and text tokens as keys and values. Its attention output goes through an FFN and a linear layer that scores each vision token as keep or remove. During training, Gumbel-Softmax turns the scores into differentiable binary decisions, an attention mask zeroes out pruned tokens without changing the token count, and an MSE loss drives the kept ratio to preset values per layer. At inference, scores are sorted and the top-k vision tokens are kept, with original positional embeddings preserved so the model sees the same distribution it saw during training.

What would settle it

Compare LVPruning at a kept ratio of 0.5 against random pruning of vision tokens at the same ratio on the nine benchmarks; if random pruning matches LVPruning's accuracy, the language-guided scoring is not what preserves performance. A complementary test: use a task whose correct answer depends on a tiny image region and see whether early pruning discards that region and causes a far larger loss than the reported ~0.45% average.

Watch

Extended reading notes

Core claim

The paper's central claim is that vision tokens can be scored by how much they matter to the language context through small cross-attention modules, and that pruning the lowest-scoring tokens at a few fixed layers preserves accuracy almost completely. On LLaVA-1.5-7B, three such modules are inserted after layers 1, 8, and 16; by the middle layer as many as 90% of vision tokens are gone. Inference cost drops from 8.38 to 3.18 TFLOPs (62.1% reduction) while the average score across VQAv2, GQA, VizWiz, SciQA-IMG, TextVQA, POPE, MMBench, LLaVA-Wild, and MM-Vet falls by about 0.45% relative to the unpruned model. The authors further show that the retention ratio can be adjusted at inference time without retraining, and that the pruned model often beats Q-former-based baselines that already use many fewer tokens.

Load-bearing premise

The load-bearing assumption is that the language tokens available at each pruning layer—the prompt and any generated tokens so far—are sufficient to identify every vision token the final answer will need, even when a small visual detail only becomes relevant deep into generation.

Editorial extensions

If this is right

Deploying a frozen MLLM becomes much cheaper: at the main setting, inference drops from 8.38 to 3.18 TFLOPs, making resource-constrained environments far more feasible.
Because the kept ratio is an inference-time knob, one model can trade a few accuracy points for large speedups on demand, with no retraining.
The base model parameters are never modified, so LVPruning can be added to an existing MLLM and later removed cleanly.
Most vision tokens disappear before the middle of the network, so the compute savings apply across the majority of layers, not just near the output.

Reading between the lines

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

If the language-guided scoring is as effective as the paper reports, a large fraction of visual tokens in LLaVA-1.5 is redundant for text-conditioned tasks; a natural but untested extension would be to apply the same scoring to higher-resolution images or video frames, where redundancy is likely even larger.
The method relies on the text available at early layers being sufficient to know which visual details the entire response will need; one test not reported is whether the pruned tokens differ when the same image is paired with different prompts, or when longer generation changes what later text requires.
A comparison against random pruning at the same kept ratios would isolate whether the language guidance, rather than mere token removal, is what preserves accuracy; the paper does not include that baseline.

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.

Desk Editor's Note

read the letter

Quick take: LVPruning is a clean, simple method for cutting vision tokens in LLaVA-style models, and the empirical story is plausible. But the headline numbers are not fully backed by the tables, the FLOPs accounting rests on a dummy input and a single forward pass, and the paper omits the obvious MLLM pruning baselines. The core idea deserves a referee, but the claims need tightening.

What's new: the cross-attention scoring where vision tokens query language tokens is a sensible extension of DynamicViT to multimodal inputs, and freezing the base model while training only the inserted modules makes it easy to apply or remove. The adjustable inference ratios without retraining are a nice practical feature. The benchmark coverage is broad, and the results at ρ=0.5 mostly show small deltas, even some gains on VizWiz and LLaVA-Wild.

Soft spots, in rough order of weight. First, the 0.45% average loss in the abstract and conclusion does not match any obvious aggregation of Tables 1-2. For ρ=0.5, the mean delta across the twelve reported metrics is about -0.3, and the mean absolute delta is about 0.68. For ρ=0.6, the deltas are mostly positive. The paper never states which configuration and which averaging produced 0.45%. That needs to be pinned down.

Second, the FLOPs numbers. Everything is computed on a dummy input of one image and 30 text tokens, with no generation length. During autoregressive decoding, the vision tokens appear only in the prompt, so the 62.1% reduction presumably applies to the prompt-processing phase, not the whole response. The paper should say so.

Third, and this is the load-bearing assumption: the cross-attention scorer is trained with the ground-truth answer tokens in the language sequence. There is no causal mask inside the decision module, so vision queries can attend to answer-token identities that do not exist at inference. No experiment separates this privileged training signal from prompt-only scoring, so the claim that pruning based on the prompt alone preserves accuracy is not yet established. The paper does not mention this mismatch anywhere.

Fourth, the missing baselines. FastV and PyramidDrop already prune vision tokens in MLLMs; the paper cites only single-modal pruning. A comparison is needed for the novelty claim to hold.

Minor: Eq. 11 defines an MSE loss but the surrounding text describes a Huber loss with β=0.5, which is a leftover edit. The Section 7 limitations are generic and do not address any of the above.

Who this is for: people working on MLLM inference efficiency. It's a useful data point, but not a fundamental advance. Deserves a serious referee; I would send it out but require a major revision that addresses the scoring mismatch and the baseline comparisons.

Referee Report

4 major / 5 minor

Summary. The paper proposes LVPruning, a method to prune vision tokens in MLLMs by inserting lightweight cross-attention decision modules into specific layers of LLaVA-1.5. The modules use vision tokens as queries and language tokens as keys/values to predict keep/remove scores; during training the decisions are applied via attention masks with Gumbel-Softmax and an MSE ratio loss, while at inference the scores are sorted and the top-k tokens are physically removed. The authors report a 62.1% reduction in inference TFLOPs at a 50% token retention ratio with roughly 0.45% average performance loss across nine multimodal benchmarks, and compare favorably against Q-former–based models.

Significance. If the central claims hold, the method is attractive: it requires no modification of the base MLLM parameters, trains only the small inserted modules, and achieves a large compute reduction with modest accuracy change. The language-guided pruning mechanism is a sensible extension of DynamicViT to multimodal settings and could be useful for deployment. However, the load-bearing claims rest on experimental details that are currently under-specified, especially the inference protocol and the exact averaging used for the headline 0.45% loss figure.

major comments (4)

[§3.3, §7, Tables 1–2] The paper never states which language tokens are available to the cross-attention decision modules at inference time. In training, the language-token set includes the ground-truth answer tokens (Eqs. 1–3), so the scorer can attend to future-answer identities. At inference, when the KV cache is used, only prompt tokens are present at the first pruning layer and only prompt plus already generated tokens at later layers. Since the method's rationale is that language context identifies useful vision tokens, the reported 0.45% loss cannot be interpreted without knowing whether the inference protocol matches this setup. The authors should specify the exact generation protocol and, ideally, provide an ablation training the decision modules with only prompt tokens (or comparing prompt-only vs. answer-token training) to verify that the method does not rely on privileged information.
[Abstract, Tables 1–2, §6] The claimed '0.45% average performance loss' at a 50% token retention ratio is not directly supported by the reported per-benchmark numbers. Computing the average signed delta over the twelve reported columns gives approximately −0.3 points, while averaging per-benchmark (collapsing the POPE and MMBench splits) gives approximately −0.3 to −0.9 points, and the relative percentage changes average to roughly +0.3%. None of these straightforward aggregations yields 0.45%. The authors must define precisely how the average is computed (e.g., mean relative drop over the nine benchmarks, or mean absolute drop) and add an average row to the results tables.
[§3.2, Eqs. (11)–(12)] Section 3.2 contains a leftover description of a Huber loss with threshold β: the sentence 'where δ(Dls, ρs) is the Huber loss, β is a threshold that determines the loss function used. We set β = 0.5' appears immediately after an equation that is a plain MSE loss with no δ or β. In addition, the loss weights λ_causal and λ_ratio in Eq. (12) are never reported, even though they control the balance between task performance and the pruning-ratio constraint. These values are necessary for reproducibility and for understanding whether the ratio loss dominates the objective.
[§4.1, Figure 4, Appendix B] All TFLOPs are computed with a single dummy input of 1 image and 30 text tokens. This is not representative of the varying prompt lengths in the nine benchmarks, and the text states that the inserted decision modules add 0.71 TFLOPs without clarifying whether this overhead is included in the reported 3.18 TFLOPs for LVPruning at ρ=0.5. The authors should describe the FLOPs methodology in more detail, including whether a KV-cache is assumed and whether the pruning decisions are applied once during prefill or at every decoding step.

minor comments (5)

[§3.3, Eq. (13)] Equation (13) uses ρs for the inference kept ratio, but the paper defines inference ratios as \\hat P = [\\hat ρ_1, ..., \\hat ρ_S]. The notation should be made consistent.
[Figure 3 and Figure 4 captions] The rendered captions show 'LVPruning ( =0.5)' with a missing ρ symbol; this appears to be a LaTeX rendering issue and should be fixed.
[Table 1] The asterisk on LLaVA-1.5's SQA-IMG score (69.4*) is never explained in the text or caption.
[§2, References] The IDEFICS model is cited to the OBELICS paper (Laurençon et al., 2023), but the model itself is described in a separate technical report; the citation should be corrected or supplemented.
[§4.2] The paper states it removes all training entries without image inputs, but does not report the exact number of remaining samples; the approximate 620k number should be made precise, or a reference to the dataset version should be given.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: retention ratios are hand-set targets, benchmarks are externally measured, and the DynamicViT-derived machinery is external prior art rather than a self-citation loop.

full rationale

LVPruning's claimed derivation chain is not circular. The token retention ratios are hand-set engineering targets (Section 4.1 sets P=[ρ,rho-0.2,rho-0.4] with rho=0.5), and the 62.1% TFLOPs figure is an arithmetic consequence of those ratios under the stated dummy-input protocol, not a predicted outcome. The benchmark numbers are externally measured on public datasets, so the 'average performance loss of 0.45%' is an empirical observation rather than a quantity constructed from the training loss. The Gumbel-Softmax and attention-mask machinery is inherited from DynamicViT (Rao et al., 2021), which is external prior work, and no load-bearing uniqueness theorem or self-citation is invoked. The one substantive concern is a train/inference context mismatch: Eqs. 1-3 compute pruning scores by cross-attending to all language tokens without a causal mask, and during training H_IT contains ground-truth answer tokens, while at the first autoregressive decoding step only the prompt is available; Section 7 does not acknowledge this gap. This is a correctness/verification risk about whether the trained scorer transfers to prompt-only inference, not an equation-level circularity, so it does not raise the circularity score.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The central claim rests on hand-chosen retention schedules and inference ratios, an unstated train/inference consistency assumption, and omitted training hyperparameters. No new physical entities are postulated.

free parameters (5)

token retention ratio schedule P = [0.5, 0.3, 0.1] for training with rho=0.5
Chosen by hand; controls pruning aggressiveness and directly determines TFLOPs reduction.
inference kept ratios = 0.6, 0.5, 0.45
Swept for evaluation, not fitted to test sets.
number and placement of decision modules = 3 modules after layers 1, 8, 16
Chosen by design; affects where pruning happens and the reported 90% pruning by the middle layer.
loss weights lambda_causal and lambda_ratio
Hyperparameters of the training objective; values are not reported in the paper.
beta threshold = 0.5
Mentioned in Section 3.2 as the Huber threshold but not used in equation (11); appears to be a leftover from a previous formulation.

assumptions (3)

domain assumption Cross-attention scores between vision and language tokens indicate vision token usefulness.
Central to the pruning decision; if text tokens cannot reliably identify necessary image regions, pruning will remove needed information.
domain assumption Masked attention training (Eqs. 7-9) transfers to hard token removal at inference.
The model is trained with soft masks and evaluated with hard indexing; the paper provides no analysis of this mismatch.
standard math Gumbel-Softmax yields approximately one-hot decisions that can be used in a differentiable pipeline.
Standard technique from Jang et al. / DynamicViT.

how reviews work

0 comments

Cite this review

Pith. "Pith review of LVPruning: An Effective yet Simple Language-Guided Vision Token Pruning Approach for Multi-modal Large Language Models." pith.science (2026). https://pith.science/paper/Y3MPM5MQ

@misc{pith2026250113652,
  author       = {Pith},
  title        = {Pith review of: LVPruning: An Effective yet Simple Language-Guided Vision Token Pruning Approach for Multi-modal Large Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Y3MPM5MQ}},
  note         = {Machine review of arXiv:2501.13652}
}

read the original abstract

Multi-modal Large Language Models (MLLMs) have achieved remarkable success by integrating visual and textual modalities. However, they incur significant computational overhead due to the large number of vision tokens processed, limiting their practicality in resource-constrained environments. We introduce Language-Guided Vision Token Pruning (LVPruning) for MLLMs, an effective yet simple method that significantly reduces the computational burden while preserving model performance. LVPruning employs cross-attention modules to compute the importance of vision tokens based on their interaction with language tokens, determining which to prune. Importantly, LVPruning can be integrated without modifying the original MLLM parameters, which makes LVPruning simple to apply or remove. Our experiments show that LVPruning can effectively reduce up to 90% of vision tokens by the middle layer of LLaVA-1.5, resulting in a 62.1% decrease in inference Tera Floating-Point Operations Per Second (TFLOPs), with an average performance loss of just 0.45% across nine multi-modal benchmarks.

Figures

Figures reproduced from arXiv: 2501.13652 by the authors.

**Figure 1.** LVPruning can reduce 62.1% of inference TFLOPs for LLaVA-1.5-7B with marginal performance loss across nine multi-modal benchmarks. *All TFLOPs reported in this paper are computed using a dummy input consisting of 1 image and 30 text tokens. focus on salient features while ignoring irrelevant details, allowing for highly efficient visual perception (Treisman, 1988). Inspired by this, there is a growing need to devel… view at source ↗

**Figure 2.** Overall Framework Architecture. LVPruning modules are incorporated into specific layers of an MLLM, where vision tokens serve as queries and language tokens act as keys and values. A pruning decision is predicted for each vision token. The operation denoted by ⊗ applies these decisions—serving as attention masking during training and token removal via indexing during inference. an image input is divided into patc… view at source ↗

**Figure 4.** Comparison of Inference FLOPs of LLaVA1.5-7B (Liu et al., 2023a) and LVPruning with different vision token kept ratio ρ. evaluating both computational savings and performance trade-off. Generally speaking, [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗

Figures from the paper (2 more)

**Figure 5.** Figure 5: The relationship between inference TFLOPs [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

**Figure 6.** Figure 6: TFLOPs of LVPruning with different values [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 20 canonical work pages

[1]

Jinze Bai, Shuai Bai, Shusheng Yang, Shijie Wang, Sinan Tan, Peng Wang, Junyang Lin, Chang Zhou, and Jingren Zhou. 2023. https://openreview.net/forum?id=qrGjFJVl3m Qwen-vl: A versatile vision-language model for understanding

work page 2023
[2]

Daniel Bolya, Cheng - Yang Fu, Xiaoliang Dai, Peizhao Zhang, Christoph Feichtenhofer, and Judy Hoffman. 2023. Token merging: Your vit but faster. In ICLR . OpenReview.net

work page 2023
[3]

Mengzhao Chen, Wenqi Shao, Peng Xu, Mingbao Lin, Kaipeng Zhang, Fei Chao, Rongrong Ji, Yu Qiao, and Ping Luo. 2023. Diffrate : Differentiable compression rate for efficient vision transformers. In ICCV , pages 17118--17128. IEEE

work page 2023
[4]

Wenliang Dai, Junnan Li, Dongxu Li, Anthony Meng Huat Tiong, Junqi Zhao, Weisheng Wang, Boyang Li, Pascale Fung, and Steven C. H. Hoi. 2023. Instructblip: Towards general-purpose vision-language models with instruction tuning. In NeurIPS

work page 2023
[5]

Yunchao Gong, Liu Liu, Ming Yang, and Lubomir D. Bourdev. 2014. Compressing deep convolutional networks using vector quantization. CoRR, abs/1412.6115

arXiv 2014
[6]

Yash Goyal, Tejas Khot, Douglas Summers - Stay, Dhruv Batra, and Devi Parikh. 2017. Making the V in VQA matter: Elevating the role of image understanding in visual question answering. In CVPR , pages 6325--6334. IEEE Computer Society

work page 2017
[7]

Stangl, Anhong Guo, Chi Lin, Kristen Grauman, Jiebo Luo, and Jeffrey P

Danna Gurari, Qing Li, Abigale J. Stangl, Anhong Guo, Chi Lin, Kristen Grauman, Jiebo Luo, and Jeffrey P. Bigham. 2018. Vizwiz grand challenge: Answering visual questions from blind people. In CVPR , pages 3608--3617. Computer Vision Foundation / IEEE Computer Society

work page 2018
[8]

Hinton, Oriol Vinyals, and Jeffrey Dean

Geoffrey E. Hinton, Oriol Vinyals, and Jeffrey Dean. 2015. Distilling the knowledge in a neural network. CoRR, abs/1503.02531

arXiv 2015

Show all 31 references

[9]

Hudson and Christopher D

Drew A. Hudson and Christopher D. Manning. 2019. GQA: A new dataset for real-world visual reasoning and compositional question answering. In CVPR , pages 6700--6709. Computer Vision Foundation / IEEE

2019
[10]

Rush, Douwe Kiela, Matthieu Cord, and Victor Sanh

Hugo Lauren c on, Lucile Saulnier, L \' e o Tronchon, Stas Bekman, Amanpreet Singh, Anton Lozhkov, Thomas Wang, Siddharth Karamcheti, Alexander M. Rush, Douwe Kiela, Matthieu Cord, and Victor Sanh. 2023. OBELICS: an open web-scale filtered dataset of interleaved image-text doc...

2023
[11]

Junnan Li, Dongxu Li, Silvio Savarese, and Steven C. H. Hoi. 2023 a . BLIP-2: bootstrapping language-image pre-training with frozen image encoders and large language models. In ICML , volume 202 of Proceedings of Machine Learning Research, pages 19730--19742. PMLR

2023
[12]

Yifan Li, Yifan Du, Kun Zhou, Jinpeng Wang, Wayne Xin Zhao, and Ji - Rong Wen. 2023 b . Evaluating object hallucination in large vision-language models. In EMNLP , pages 292--305. Association for Computational Linguistics

2023
[13]

Youwei Liang, Chongjian Ge, Zhan Tong, Yibing Song, Jue Wang, and Pengtao Xie. 2022. Not all patches are what you need: Expediting vision transformers via token reorganizations. CoRR, abs/2202.07800

2022 arXiv
[14]

Haotian Liu, Chunyuan Li, Yuheng Li, and Yong Jae Lee. 2023 a . Improved baselines with visual instruction tuning. CoRR, abs/2310.03744

2023 arXiv
[15]

Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. 2023 b . http://papers.nips.cc/paper\_files/paper/2023/hash/6dcf277ea32ce3288914faf369fe6de0-Abstract-Conference.html Visual instruction tuning . In Advances in Neural Information Processing Systems 36: Annual Conference...

2023
[16]

Yuan Liu, Haodong Duan, Yuanhan Zhang, Bo Li, Songyang Zhang, Wangbo Zhao, Yike Yuan, Jiaqi Wang, Conghui He, Ziwei Liu, Kai Chen, and Dahua Lin. 2023 c . Mmbench: Is your multi-modal model an all-around player? CoRR, abs/2307.06281

2023 arXiv
[17]

Pan Lu, Swaroop Mishra, Tanglin Xia, Liang Qiu, Kai - Wei Chang, Song - Chun Zhu, Oyvind Tafjord, Peter Clark, and Ashwin Kalyan. 2022. Learn to explain: Multimodal reasoning via thought chains for science question answering. In NeurIPS

2022
[18]

Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. 2021. http://proceedings.mlr.press/v139/radford21a.html Learning transferable visual models...

2021
[19]

Yongming Rao, Wenliang Zhao, Benlin Liu, Jiwen Lu, Jie Zhou, and Cho - Jui Hsieh. 2021. Dynamicvit: Efficient vision transformers with dynamic token sparsification. In NeurIPS, pages 13937--13949

2021
[20]

Victor Sanh, Lysandre Debut, Julien Chaumond, and Thomas Wolf. 2019. Distilbert, a distilled version of BERT: smaller, faster, cheaper and lighter. CoRR, abs/1910.01108

2019 arXiv
[21]

Amanpreet Singh, Vivek Natarajan, Meet Shah, Yu Jiang, Xinlei Chen, Dhruv Batra, Devi Parikh, and Marcus Rohrbach. 2019. Towards VQA models that can read. In CVPR , pages 8317--8326. Computer Vision Foundation / IEEE

2019
[22]

Yizheng Sun, Hao Li, Chenghua Lin, and Riza Batista - Navarro. 2024. Lanvikd: Cross-modal language-vision knowledge distillation for egocentric action recognition. In HAII5.0@ECAI, volume 3765 of CEUR Workshop Proceedings . CEUR-WS.org

2024
[23]

Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, Dan Bikel, Lukas Blecher, Cristian Canton - Ferrer, Moya Chen, Guillem Cucurull, David Esiobu, Jude Fernandes, Jeremy Fu,...

2023 doi arXiv
[24]

Anne Treisman. 1988. https://doi.org/10.1080/02724988843000104 Features and objects: The fourteenth bartlett memorial lecture . The Quarterly Journal of Experimental Psychology Section A, 40(2):201--237. PMID: 3406448

1988 doi
[25]

Gomez, Lukasz Kaiser, and Illia Polosukhin

Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. 2017. https://proceedings.neurips.cc/paper/2017/hash/3f5ee243547dee91fbd053c1c4a845aa-Abstract.html Attention is all you need . In Advances in Neural I...

2017
[26]

Kuan Wang, Zhijian Liu, Yujun Lin, Ji Lin, and Song Han. 2019. HAQ: hardware-aware automated quantization with mixed precision. In CVPR , pages 8612--8620. Computer Vision Foundation / IEEE

2019
[27]

Wenhui Wang, Furu Wei, Li Dong, Hangbo Bao, Nan Yang, and Ming Zhou. 2020. Minilm: Deep self-attention distillation for task-agnostic compression of pre-trained transformers. In NeurIPS

2020
[28]

Weihao Yu, Zhengyuan Yang, Linjie Li, Jianfeng Wang, Kevin Lin, Zicheng Liu, Xinchao Wang, and Lijuan Wang. 2024. Mm-vet: Evaluating large multimodal models for integrated capabilities. In ICML . OpenReview.net

2024
[29]

Deyao Zhu, Jun Chen, Xiaoqian Shen, Xiang Li, and Mohamed Elhoseiny. 2024. Minigpt-4: Enhancing vision-language understanding with advanced large language models. In ICLR . OpenReview.net

2024
[30]

online" 'onlinestring :=

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list...
[31]

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 gl...

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [1]

Jinze Bai, Shuai Bai, Shusheng Yang, Shijie Wang, Sinan Tan, Peng Wang, Junyang Lin, Chang Zhou, and Jingren Zhou. 2023. https://openreview.net/forum?id=qrGjFJVl3m Qwen-vl: A versatile vision-language model for understanding

work page 2023

[2] [2]

Daniel Bolya, Cheng - Yang Fu, Xiaoliang Dai, Peizhao Zhang, Christoph Feichtenhofer, and Judy Hoffman. 2023. Token merging: Your vit but faster. In ICLR . OpenReview.net

work page 2023

[3] [3]

Mengzhao Chen, Wenqi Shao, Peng Xu, Mingbao Lin, Kaipeng Zhang, Fei Chao, Rongrong Ji, Yu Qiao, and Ping Luo. 2023. Diffrate : Differentiable compression rate for efficient vision transformers. In ICCV , pages 17118--17128. IEEE

work page 2023

[4] [4]

Wenliang Dai, Junnan Li, Dongxu Li, Anthony Meng Huat Tiong, Junqi Zhao, Weisheng Wang, Boyang Li, Pascale Fung, and Steven C. H. Hoi. 2023. Instructblip: Towards general-purpose vision-language models with instruction tuning. In NeurIPS

work page 2023

[5] [5]

Yunchao Gong, Liu Liu, Ming Yang, and Lubomir D. Bourdev. 2014. Compressing deep convolutional networks using vector quantization. CoRR, abs/1412.6115

arXiv 2014

[6] [6]

Yash Goyal, Tejas Khot, Douglas Summers - Stay, Dhruv Batra, and Devi Parikh. 2017. Making the V in VQA matter: Elevating the role of image understanding in visual question answering. In CVPR , pages 6325--6334. IEEE Computer Society

work page 2017

[7] [7]

Stangl, Anhong Guo, Chi Lin, Kristen Grauman, Jiebo Luo, and Jeffrey P

Danna Gurari, Qing Li, Abigale J. Stangl, Anhong Guo, Chi Lin, Kristen Grauman, Jiebo Luo, and Jeffrey P. Bigham. 2018. Vizwiz grand challenge: Answering visual questions from blind people. In CVPR , pages 3608--3617. Computer Vision Foundation / IEEE Computer Society

work page 2018

[8] [8]

Hinton, Oriol Vinyals, and Jeffrey Dean

Geoffrey E. Hinton, Oriol Vinyals, and Jeffrey Dean. 2015. Distilling the knowledge in a neural network. CoRR, abs/1503.02531

arXiv 2015