REVIEW 4 major objections 5 minor 62 references
Foveated Probes Recover Localized Binding Information in Vision Foundation Models
T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash
Pith's one-line read Frozen patch tokens keep color–shape binding that global readouts discard.
desk verdict The readout-vs-representation distinction is solid and well-evidenced, but the paper should verify where its foveated readout actually looks on the synthetic tasks before claiming localization. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The central objects are three readouts applied to frozen patch tokens: global average pooling (uniform mean), the pretrained summary token, and a foveated readout that attention-pools tokens using a single query—a learned vector for vision-only tasks, or a linear map from the frozen question embedding plus a learned temperature for vision-language tasks. The oracle readout, Eq. (1), is the mean over patch tokens overlapping the annotated target region, and it serves as an upper bound on what the frozen tokens carry. The mechanism is the counterfactual nuisance-to-signal ratio (NSR), which measures how much a readout's embedding changes under label-changing target edits relative to label-preserving nuisance edits, together with the dilution identity $\Delta z_{\text{gap}} \approx \frac{M}{N}\Delta x_{\text{tar}}$: global averaging attenuates the localized representation shift by the fraction $M/N$ of affected tokens, while foveated attention preserves it in proportion to the attention mass on the target.
What would settle it
If one re-runs the GQA-derived experiment with masks jittered or shifted by a single patch and oracle paired accuracy drops from 33% toward the 4% global level, the apparent token-level information would be an artifact of mask placement. Similarly, if a foveated readout trained without location supervision fails to beat global readouts on a re-annotated GQA-derived set with tight boxes, the claim that question-conditioned selection recovers localized binding would be falsified.
Extended reading notes
Core claim
Holding a frozen vision encoder fixed and changing only the readout, the paper finds that localized binding information—the conjunction of color and shape of one target object among distractors, or the color of a particular same-category instance selected by a question—is present in the final patch tokens. The pretrained summary token and global average pooling both collapse that signal: they are near chance on paired counterfactuals (3.2±0.1% and 3.5±0.2% on CSB) and their embeddings move more under label-preserving nuisance edits than under label-changing target edits, with a counterfactual nuisance-to-signal ratio near 7–15 versus roughly 0.3 for selective readouts. A single-query attention readout trained from the task loss alone, with no location supervision, reaches 93.5±0.6% paired accuracy, and an oracle pooling tokens inside the annotated target region reaches 99.3%, showing that the ceiling is set by the frozen tokens, not the probe. On GQA-derived natural images, question-conditioned foveation reaches 17.0±0.7% paired accuracy versus 3.8–4.3% for global visual readouts and 33.0% for the oracle. The paper's central claim is that readout-level spatial blindness, not representation failure, explains these global-probe failures.
Load-bearing premise
The oracle's mask in Eq. (1) is assumed to select exactly the patch tokens that determine the label; if the GQA-derived boxes are misaligned or include non-target content, the oracle upper bound and the fovea's gap to it no longer cleanly measure the information in the frozen tokens.
Editorial extensions
If this is right
- Global readouts do not certify representational absence: a near-chance score from a summary token or global average pooling on a localized task should be re-checked with a readout that can select tokens before pooling.
- A lightweight readout trained without location supervision can recover most of the oracle-accessible binding signal, so task-specific spatial selection need not require fine-tuning the encoder.
- The counterfactual nuisance-to-signal ratio gives a representation-level predictor of readout quality: readouts with NSR near or above 1 can be expected to fail as clutter grows, while NSR below roughly 0.3 tracks near-oracle accuracy.
- Question-conditioned foveation improves paired localized color accuracy on natural images by roughly four times over question-independent global vectors, and the same ordering holds across CLIP, SigLIP, and SigLIP 2 with both linear and MLP probes.
Reading between the lines
- A testable extension: train the foveated readout with oracle masks as soft supervision and measure how much of the remaining gap to the oracle closes.
- A benchmark-level consequence the authors leave implicit is that compositionality scores obtained through global embeddings may need to be recomputed with selective readouts before they are interpreted as showing representational failure.
- The GQA-derived oracle ceiling of 33% is far below the synthetic ceiling; a re-annotation pass with tighter boxes and unambiguous color labels would show how much of that gap is annotation noise rather than token-level absence.
- The diagnostic could be carried to mid-layer tokens and to relational predicates such as above/below and left/right to test whether readout-level blindness is a general interface phenomenon rather than specific to attribute binding.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper investigates whether apparent spatial blindness in frozen vision foundation models reflects an absence of localized information in the representation or a loss caused by the global readout interface. Keeping the encoder fixed, it compares global average pooling (GAP), the pretrained summary token, a foveated attention readout with a learned or question-conditioned query, and a non-deployable oracle readout that selects patch tokens within an annotated target region. The evaluation covers three tasks: a synthetic color-shape binding task under clutter (CSB), a color-free crowded shape-detection variant (CSD), and a GQA-derived paired localized-color task (GQA*). The main results are that global readouts collapse on paired-counterfactual accuracy under clutter (3.5% for GAP versus 93.5% for the foveated readout and 99.3% for the oracle on CSB), and that on GQA* question-conditioned foveation (17.0%) substantially outperforms global readouts (3.8-4.3%) while the oracle reaches 33.0%. A counterfactual nuisance-to-signal ratio (NSR) is introduced and shown to be high for global readouts and low for selective readouts. The paper concludes that apparent spatial blindness can arise from the global embedding interface rather than from the absence of localized binding information in frozen patch tokens.
Significance. If the central claim holds, the paper makes a valuable conceptual and methodological contribution: it separates readout-level information loss from representational absence, provides a quantitative NSR diagnostic, and demonstrates with an oracle control that patch tokens retain localized binding information that global embeddings discard. The experimental execution is a strength: the encoder is frozen, readout and probe architectures are matched across conditions, results are averaged over five seeds, three encoders are tested, both linear and MLP probes are used, and the counterfactual pairs are matched. The oracle result makes the information-presence claim robust, and the consistent ordering across configurations supports the interface-level interpretation. The main uncertainties concern the mechanistic interpretation of the foveated readout on the synthetic tasks and the consistency of the natural-image oracle procedure; these are fixable but currently load-bearing for parts of the paper's claims.
major comments (4)
- [Eq. (1) vs Appendix 'GQA-derived color binding (GQA*)'] The oracle readout defined in Eq. (1) is a masked mean over patch tokens of the full image, but the dataset-construction appendix states that 'The oracle input is the target's ground-truth box cropped from the resized image and stretched to 384×384.' These are different operations: Eq. (1) preserves the original image context and token grid, whereas a crop input changes resolution and removes all other objects. The reported 33.0% GQA* oracle paired accuracy is therefore not an upper bound for information accessible from the frozen tokens of the full image, and the statement that 'the frozen spatial representation contains additional object-specific color information' is not directly supported by this comparison. Please reconcile the two descriptions; if the crop procedure was used, report a masked-token oracle on the full image as well, or reinterpret the foveation-to-oracle gap accordingly.
- ['Attention Concentration Diagnostics'; 'Experimental Protocol' / Figure 4] Direct evidence about where the foveated readout looks is reported only for the GQA*-derived task: EffPatches=5.2±1.2, MaxAttn=0.621±0.035, with a learned temperature of 14.2±3.3. The synthetic CSB/CSD foveated readout uses a single learned query with no temperature and no attention diagnostics. Because the headline 93.5% CSB result is obtained in this setting, the claim that the readout is genuinely 'foveated' on the target, rather than suppressing the white background or broadly weighting foreground tokens, is currently unverified. The lure-invariance and NSR results are suggestive but indirect. Please report attention entropy, maximum attention weight, and ideally the overlap between attended patches and the target mask on CSB/CSD; if the attention is not target-concentrated, the 'localized binding' interpretation of the foveated readout should be revised.
- ['Foveated Readout' and 'Experimental Protocol'] The comparison between foveated and global readouts is not capacity-matched: the foveated readout adds a learned query q (and, on GQA*, a learned W_q and temperature), while GAP and the summary token are fixed operations on top of the same probe architecture. The large accuracy gap could therefore be attributed, at least in part, to the extra trainable parameters rather than to spatial selection. The oracle result already establishes that localized information exists, but to support the claim that the foveated protocol succeeds because of selective pooling, the authors should include a control readout with comparable added parameters but without spatial selection, for example a learned per-token weighting applied before global averaging or an attention readout whose query is not trained.
- [Eqs. (5)-(7) and Appendix 'Color-shape binding (CSB)'] The counterfactual pairs used for the NSR analysis are generated by swapping the target's color or shape with a matched non-target object, so the matched partner also changes appearance. Eq. (6) assumes that 'token displacements outside the target region are negligible,' which is violated by this partner displacement. The illustrative attenuation factor (M/N)^2 is thus not a complete description of the GAP representation change, since the partner edit contributes an additional displacement outside the target mask. Please either revise the derivation to include both changed token sets or restrict the illustrative claim to the target-token component and note the partner contribution separately.
minor comments (5)
- ['Main Readout Comparison and Counterfactual Sensitivity' / Table 1] The main text states that 'all readouts are near ceiling (≥99.6%) on the clean no-distractor split,' but Appendix Table 1 reports CLIP GAP clean accuracy of 52.8% (linear) and 45.9% (MLP) and CLIP Summary MLP clean accuracy of 96.9% on CSB. This claim should be qualified to the configurations shown in Figure 4 or corrected.
- ['GQA-derived localized color binding (GQA*)'] The question-only baseline is described as 2.3% paired accuracy versus a 1% uniform-chance level, but the GQA* label distribution is imbalanced and paired chance is not estimated under the empirical label prior. Reporting chance under the empirical class distribution would clarify how much of the 2.3% reflects text priors.
- ['Foveated Readout'] Because the synthetic-task query is content-based and shared across images rather than position-based, the term 'foveated' may overstate the degree of spatial localization. Consider renaming the readout 'attention readout' or adding a clarifying sentence that 'foveation' refers to attention pooling over tokens, not to a spatially anchored fovea.
- [References / Related Work] The reference for Subramanian et al. should be capitalized as 'ReCLIP' and the related-work sentence 'Reclip: A strong zero-shot baseline for referring expression comprehension' should be harmonized with the citation format.
- [Appendix 'Computing Infrastructure'] No code or data release is mentioned; for a diagnostic protocol, providing code for the foveated readout and NSR computation would materially improve reproducibility.
Circularity Check
No circularity: the readout comparisons and NSR diagnostic are self-contained empirical measurements.
full rationale
The paper's central claim is that frozen patch tokens retain localized binding information that global readouts discard, and the evidence is a direct experimental comparison of readouts on a fixed frozen encoder. The foveated readout is trained only on task labels with no location supervision and is evaluated on held-out counterfactual pairs; its 93.5% paired accuracy is a generalization measurement, not a fitted input renamed as a prediction. The oracle readout (Eq. 1) is an annotation-defined diagnostic upper bound, not a learned predictor, so it independently supports the information-presence claim. The NSR metric (Eqs. 2-4) is a descriptive statistic computed on readout embeddings; it is not fitted to the accuracy results, and the signal-dilution derivation (Eqs. 5-7) is an explanatory illustration rather than an experimental input. All references are to external work; there is no load-bearing self-citation chain, no imported uniqueness theorem, and no ansatz smuggled in via the authors' own prior results. The paper explicitly acknowledges that the oracle's absolute level on GQA* partly reflects noisy scene-graph boxes and color labels, which is a stated limitation rather than a circular step. The only notable evidentiary gap is that attention-concentration diagnostics are reported for the GQA* task but not directly for the synthetic CSB/CSD tasks; this bears on the strength of the mechanism claim, not on whether any result reduces by construction to its inputs.
Assumptions & free parameters
free parameters (4)
- Foveated query vector q (CSB/CSD) =
learned, R^1152 for SigLIP, init N(0,0.02^2)
- Question-conditioned query map Wq (GQA*) =
learned d_text x d_patch matrix, e.g., 768x1024 for CLIP
- Softmax temperature T (GQA*) =
learned, converged to 14.2 +/- 3.3 from init sqrt(1152)
- NSR denominator epsilon =
1e-8
assumptions (5)
- domain assumption Final patch tokens of the frozen encoder contain the attribute information needed for the tested tasks when read out selectively.
- domain assumption Constructed counterfactual pairs change only the label-defining factor while preserving nuisance statistics, so paired accuracy measures sensitivity to the target binding rather than to global image statistics.
- domain assumption Oracle target-region masks derived from annotations correctly cover the patch tokens that determine the label.
- standard math Z-scoring readout coordinates on the base training split is an appropriate standardization for comparing representation change across readouts of different scales.
- standard math Softmax attention pooling yields a differentiable weighted average of frozen patch tokens.
Cite this review
Pith. "Pith review of Foveated Probes Recover Localized Binding Information in Vision Foundation Models." pith.science (2026). https://pith.science/paper/WT5ARY24
@misc{pith2026260800726,
author = {Pith},
title = {Pith review of: Foveated Probes Recover Localized Binding Information in Vision Foundation Models},
year = {2026},
howpublished = {\url{https://pith.science/paper/WT5ARY24}},
note = {Machine review of arXiv:2608.00726}
}
read the original abstract
Frozen vision foundation models are commonly evaluated through a single global image embedding, but this interface can conflate missing information with information lost at readout time. We study this distinction by keeping a pretrained vision encoder frozen and varying only the readout applied to its final patch tokens. We compare standard global readouts against a lightweight foveated readout, which attention-pools patch tokens using a learned or question-conditioned query, and against an oracle readout with access to the annotated target region. We evaluate these interfaces on three localized binding problems: a controlled synthetic color--shape binding task under clutter, a color-free crowded shape-detection variant, and a GQA-derived natural-image task where paired questions ask for the colors of different same-category objects in the same image. Global readouts perform near perfectly when the synthetic target appears alone, but collapse under clutter and counterfactual target edits, whereas the foveated readout recovers most of the oracle-accessible signal. On the GQA-derived task, question-independent global image vectors improve only modestly over question-only priors, while question-conditioned foveation substantially improves paired localized color accuracy. A counterfactual nuisance-to-signal ratio explains the synthetic failures: global pooling dilutes localized label-changing evidence while exposing the probe to nuisance variation from irrelevant objects. These results indicate that apparent spatial blindness in frozen vision models can arise from the global embedding interface rather than from an absence of spatial information in the frozen patch tokens.
Figures
Figures from the paper (1 more)
Reference graph
Works this paper leans on
-
[1]
Information processing letters , volume=
Another efficient algorithm for convex hulls in two dimensions , author=. Information processing letters , volume=. 1979 , publisher=
work page 1979
-
[2]
arXiv preprint arXiv:2412.15115 , year =
Qwen2.5 Technical Report , author =. arXiv preprint arXiv:2412.15115 , year =
-
[3]
, author=
Lora: Low-rank adaptation of large language models. , author=. ICLR , volume=
-
[4]
Proceedings of the IEEE/CVF international conference on computer vision , pages=
Sigmoid loss for language image pre-training , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=
-
[5]
Cultural recycling of cortical maps , author=. Neuron , volume=. 2007 , publisher=
work page 2007
-
[6]
arXiv preprint arXiv:2011.04006 , year=
Long range arena: A benchmark for efficient transformers , author=. arXiv preprint arXiv:2011.04006 , year=
arXiv 2011
-
[7]
Proceedings of the IEEE conference on computer vision and pattern recognition , pages=
Clevr: A diagnostic dataset for compositional language and elementary visual reasoning , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=
-
[8]
Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
MVTec AD--A comprehensive real-world dataset for unsupervised anomaly detection , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
Show all 62 references
-
[9]
European conference on computer vision , pages=
Microsoft coco: Common objects in context , author=. European conference on computer vision , pages=. 2014 , organization=
2014
-
[10]
Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
Camouflaged object detection , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
-
[11]
Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
Towards vqa models that can read , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
-
[12]
2015 13th international conference on document analysis and recognition (ICDAR) , pages=
ICDAR 2015 competition on robust reading , author=. 2015 13th international conference on document analysis and recognition (ICDAR) , pages=. 2015 , organization=
2015
-
[13]
Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=
Spatialsense: An adversarially crowdsourced benchmark for spatial relation recognition , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=
-
[14]
International journal of computer vision , volume=
Visual genome: Connecting language and vision using crowdsourced dense image annotations , author=. International journal of computer vision , volume=. 2017 , publisher=
2017
-
[15]
Advances in neural information processing systems , volume=
A simple neural network module for relational reasoning , author=. Advances in neural information processing systems , volume=
-
[16]
Proceedings of the IEEE conference on computer vision and pattern recognition , pages=
Single-image crowd counting via multi-column convolutional neural network , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=
-
[17]
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
Learning to count everything , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
-
[18]
arXiv preprint arXiv:2010.11929 , year=
An image is worth 16x16 words: Transformers for image recognition at scale , author=. arXiv preprint arXiv:2010.11929 , year=
2010 arXiv
-
[19]
Proceedings of the IEEE conference on computer vision and pattern recognition , pages=
From image-level to pixel-level labeling with convolutional networks , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=
-
[20]
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
Frozen feature augmentation for few-shot image classification , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
-
[21]
Advances in neural information processing systems , volume=
Flamingo: a visual language model for few-shot learning , author=. Advances in neural information processing systems , volume=
-
[22]
International conference on machine learning , pages=
Set transformer: A framework for attention-based permutation-invariant neural networks , author=. International conference on machine learning , pages=. 2019 , organization=
2019
-
[23]
Advances in neural information processing systems , volume=
Do vision transformers see like convolutional neural networks? , author=. Advances in neural information processing systems , volume=
-
[24]
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
Winoground: Probing vision and language models for visio-linguistic compositionality , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
-
[25]
arXiv preprint arXiv:2210.01936 , year=
When and why vision-language models behave like bags-of-words, and what to do about it? , author=. arXiv preprint arXiv:2210.01936 , year=
-
[26]
Transactions of the Association for Computational Linguistics , volume=
Visual spatial reasoning , author=. Transactions of the Association for Computational Linguistics , volume=. 2023 , publisher=
2023
-
[27]
arXiv preprint arXiv:2204.05991 , year=
Reclip: A strong zero-shot baseline for referring expression comprehension , author=. arXiv preprint arXiv:2204.05991 , year=
-
[28]
Monthly Notices of the Royal Astronomical Society , volume=
Galaxy Zoo 2: detailed morphological classifications for 304 122 galaxies from the Sloan Digital Sky Survey , author=. Monthly Notices of the Royal Astronomical Society , volume=. 2013 , publisher=
2013
-
[29]
Psychonomic bulletin & review , volume=
Guided search 2.0 a revised model of visual search , author=. Psychonomic bulletin & review , volume=. 1994 , publisher=
1994
-
[30]
, author=
Visual search and stimulus similarity. , author=. Psychological review , volume=. 1989 , publisher=
1989
-
[31]
Cognitive psychology , volume=
A feature-integration theory of attention , author=. Cognitive psychology , volume=. 1980 , publisher=
1980
-
[32]
Cognition , volume=
The role of location indexes in spatial perception: A sketch of the FINST spatial-index model , author=. Cognition , volume=. 1989 , publisher=
1989
-
[33]
Readings in computer vision , pages=
Visual routines , author=. Readings in computer vision , pages=. 1987 , publisher=
1987
-
[34]
Cognitive psychology , volume=
The reviewing of object files: Object-specific integration of information , author=. Cognitive psychology , volume=. 1992 , publisher=
1992
-
[35]
, author=
Spatial attention and the apprehension of spatial relations. , author=. Journal of Experimental Psychology: Human Perception and Performance , volume=. 1994 , publisher=
1994
-
[36]
, author=
Subitizing: an analysis of its component processes. , author=. Journal of experimental psychology: general , volume=. 1982 , publisher=
1982
-
[37]
, author=
Why are small and large numbers enumerated differently? A limited-capacity preattentive stage in vision. , author=. Psychological review , volume=. 1994 , publisher=
1994
-
[38]
Cognitive psychology , volume=
Linguistic and conceptual control of visual spatial attention , author=. Cognitive psychology , volume=. 1995 , publisher=
1995
-
[39]
Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
Gqa: A new dataset for real-world visual reasoning and compositional question answering , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
-
[40]
Advances in Neural Information Processing Systems , volume=
Understanding the limits of vision language models through the lens of the binding problem , author=. Advances in Neural Information Processing Systems , volume=
-
[41]
Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
Eyes wide shut? exploring the visual shortcomings of multimodal llms , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
-
[42]
Findings of the Association for Computational Linguistics: EACL 2024 , pages=
Does clip bind concepts? probing compositionality in large image models , author=. Findings of the Association for Computational Linguistics: EACL 2024 , pages=
2024
-
[43]
Proceedings of the Asian Conference on Computer Vision , pages=
Vision language models are blind , author=. Proceedings of the Asian Conference on Computer Vision , pages=
-
[44]
arXiv preprint arXiv:2502.03566 , year=
Clip behaves like a bag-of-words model cross-modally but not uni-modally , author=. arXiv preprint arXiv:2502.03566 , year=
-
[45]
arXiv preprint arXiv:2504.02328 , year=
Refining CLIP's Spatial Awareness: A Visual-Centric Perspective , author=. arXiv preprint arXiv:2504.02328 , year=
-
[46]
arXiv preprint arXiv:2503.19707 , year=
Mind the gap: Benchmarking spatial reasoning in vision-language models , author=. arXiv preprint arXiv:2503.19707 , year=
-
[47]
Advances in Neural Information Processing Systems , volume=
Visual structures help visual reasoning: Addressing the binding problem in LVLMs , author=. Advances in Neural Information Processing Systems , volume=
-
[48]
arXiv preprint arXiv:2507.07102 , year=
Does Data Scaling Lead to Visual Compositional Generalization? , author=. arXiv preprint arXiv:2507.07102 , year=
-
[49]
European Conference on Computer Vision , pages=
The hard positive truth about vision-language compositionality , author=. European Conference on Computer Vision , pages=. 2024 , organization=
2024
-
[50]
International conference on machine learning , pages=
Learning transferable visual models from natural language supervision , author=. International conference on machine learning , pages=. 2021 , organization=
2021
-
[51]
arXiv preprint arXiv:2207.00221 , year=
Vl-checklist: Evaluating pre-trained vision-language models with objects, attributes and relations , author=. arXiv preprint arXiv:2207.00221 , year=
-
[52]
EMNLP , year=
What's ``up'' with vision-language models? Investigating their struggle with spatial reasoning , author=. EMNLP , year=
-
[53]
Cognitive psychology , volume=
Illusory conjunctions in the perception of objects , author=. Cognitive psychology , volume=. 1982 , publisher=
1982
-
[54]
European conference on computer vision , pages=
Extract free dense labels from clip , author=. European conference on computer vision , pages=. 2022 , organization=
2022
-
[55]
Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
Denseclip: Language-guided dense prediction with context-aware prompting , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
-
[56]
European conference on computer vision , pages=
Sclip: Rethinking self-attention for dense vision-language inference , author=. European conference on computer vision , pages=. 2024 , organization=
2024
-
[57]
International conference on machine learning , pages=
Show, attend and tell: Neural image caption generation with visual attention , author=. International conference on machine learning , pages=. 2015 , organization=
2015
-
[58]
Proceedings of the IEEE conference on computer vision and pattern recognition , pages=
Bottom-up and top-down attention for image captioning and visual question answering , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=
-
[59]
International conference on machine learning , pages=
Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models , author=. International conference on machine learning , pages=. 2023 , organization=
2023
-
[60]
Advances in neural information processing systems , volume=
Visual instruction tuning , author=. Advances in neural information processing systems , volume=
-
[61]
arXiv preprint arXiv:2502.14786 , year=
SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features , author=. arXiv preprint arXiv:2502.14786 , year=
-
[62]
International Conference on Learning Representations (ICLR) , year=
Adam: A Method for Stochastic Optimization , author=. International Conference on Learning Representations (ICLR) , year=
Reviewed August 15, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.