Pith. sign in

REVIEW 3 major objections 6 minor 90 references

Zero-shot 2D Grounding with Novel Affordance Types

T0 review · 3 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read Zero-shot grounding of novel affordance verbs works by detecting the object with the verb and scoring its subpart patches with a frozen vision-language model.

desk verdict Useful new benchmark for novel-affordance grounding, but the headline numbers rest on pair-level splits that may leak test images into validation. read the letter →

arxiv 2608.08929 v1 pith:DI7RXLUZ submitted 2026-08-09 cs.CV

Haomeng Zhang , Raymond A. Yeh This is my paper · ORCID
classification cs.CV
keywords affordancegroundingnoveltypeszero-shotgeneralization2Dvision-languagemodelsopen-vocabularysegmentationobjectsubpartpatchesNATbenchmark
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

Affordance grounding usually assumes the set of human–object interaction verbs is fixed at training time. This paper introduces the novel-affordance-types (NAT) setting, in which the test verbs are disjoint from training verbs, and argues that existing one-shot affordance methods and open-vocabulary segmentation models fail at it. The proposed training-free pipeline, AffordAnything, detects the object using the affordance verb as an open-vocabulary query and then scores refined subpart patches with a frozen vision-language model; the trainable AffordAnything+ learns only a small fusion module on top. On the re-split AGD20K-NAT benchmark, the best variant improves IoU@0.4 by 12.3 absolute percentage points over the prior state of the art, supporting the paper's claim that verb-level generalization is achievable without task-specific supervised training.

What carries the argument

The mechanism is the local-patch decomposition of object masks: each detected mask is cut into an even grid of interior patches plus a border-focused set, so affordance regions that sit on a subpart (handle, rim, edge, opening) become individual candidates. Cosine similarity between the verb's text embedding and each patch's visual embedding, softmaxed over patches, provides a training-free attention map; AffordAnything+ adds a small trainable fusion module with a branch-weight predictor, a local weight bias, and a cross-attention global decoder that are learned with one-shot supervision while the patch scoring stays grounded in the frozen model.

What would settle it

Present the pipeline with images where the same verb names no object (e.g., 'drink' on a sealed bottle with no cup) and check whether a plausible but wrong object mask is returned; if random substitute verbs produce equally good grounding masks, the affordance-specific reasoning is not doing the work. A simpler check already exists in the paper: compare the samples where the detector returns no object against a version that always returns the full-image mask, since the gap measures how much of the result rests on detection.

Watch

Extended reading notes

Core claim

The paper claims that zero-shot grounding of never-seen affordance types is possible by decomposition and scoring rather than by learning a fixed affordance vocabulary. Given an image and a novel verb, the method queries an open-vocabulary detector with the verb itself, refines each detected object mask into evenly focused and border-focused local patches, and uses cosine similarity between the verb's text embedding and each patch's visual embedding to build a soft affordance map. A global image-level branch is fused with the local branch, and in the trainable variant a lightweight decoder and learned fusion weights refine the combination. The paper reports that on AGD20K-NAT this pipeline reaches 21.3 IoU@0.4 in the training-free setting and 22.9 in the one-shot setting, against 10.6 for the strongest prior affordance model, and that the same pipeline also generalizes to novel object–novel affordance combinations.

Load-bearing premise

The pipeline assumes the affordance verb, used alone as a text query, will make the open-vocabulary detector find the object whose part is the target region; when detection fails, the local branch is silent, and the paper's appendix reports detection on only about 80% of test samples.

Editorial extensions

If this is right

  • Current closed-set affordance models do not merely underperform; they fail to transfer to unseen verbs, so new benchmarks are needed to measure verb-level generalization.
  • A training-free combination of open-vocabulary detection and patch scoring with a frozen vision-language model is enough for substantial gains, meaning the bottleneck is not task-specific training data.
  • The 12.3-point IoU@0.4 improvement on AGD20K-NAT, with consistent gains on UMD-NAT and on two additional splits, suggests the finding is stable across different ways of partitioning affordance types.
  • Because the local cue is affordance-independent, the same patch-scoring module can be reused for any new verb without retraining the backbone.

Reading between the lines

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

  • If detection is the limiting factor, then supplying the object category as an auxiliary query (for example, the object named alongside the verb) should raise the roughly 80% detection recall and lift grounding further; the paper's own failure analysis implies this test.
  • The spatial-patch decomposition is a generic inductive bias for function-bearing subregions, so the same recipe might transfer to 3D affordance grounding or part-based robotic manipulation once a 3D mask decomposition replaces the 2D patches.
  • The learned branch weight likely encodes how much a verb is tied to object identity versus location; measuring the branch weight across verbs predicts that abstract verbs like 'use' lean more on the global branch than spatial verbs like 'open'.
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

3 major / 6 minor

Summary. The paper proposes a new task, zero-shot 2D grounding with novel affordance types (NAT), and introduces NAT splits of the AGD20K and UMD datasets. The authors propose AffordAnything, a training-free pipeline that detects objects with Grounded-SAM, decomposes the object masks into local patches, and scores the patches with CLIP features against the affordance text, fusing global and local branches. They also propose AffordAnything+, a trainable extension that learns a branch-weight predictor, a local weight bias, and a global branch decoder. On the proposed benchmarks the paper reports large gains over existing affordance grounding and open-vocabulary segmentation baselines, including a 12.3% absolute IoU@0.4 improvement over OOAL on AGD20K-NAT, and it claims that existing methods fail to generalize to novel affordance types while the proposed approach succeeds.

Significance. If the empirical claims hold, this paper would fill a genuine gap: prior 2D affordance grounding almost exclusively evaluates novel object categories with fixed affordance sets, while the NAT setting is closer to open-ended instructions in real applications. The proposed method is simple, modular, and grounded in an interpretable intuition about object subparts, and the appendix contains useful ablations and detection-recall analysis. The significance is, however, currently conditional on the benchmark being clean: the pair-level splitting described in Appendix A1 creates a real risk of image leakage between validation and test, and the central numbers in Tables 1-6 are exactly what would be affected. The paper also ships no code for the new splits, so the leak cannot be verified or ruled out by inspection.

major comments (3)
  1. [Appendix A1, Sec. 5.1, Tables 1-3] The NAT splits are constructed at the image-affordance-pair level, not the image level. Appendix A1 states that 'For the validation and test data, we use the entire AGD20K test set in the Seen setting and further split it following the divided validation and test affordance types.' Because a single image can carry annotations for multiple affordance types, the same image can appear in both the validation set and the test set (and potentially in the training set) under this procedure. The main text only asserts that affordance types are disjoint across splits, which says nothing about image overlap. Since Appendix A2 states that AffordAnything+ selects the best model based on the validation set, shared images between validation and test mean that model selection can exploit the visual content of test images. This directly affects the credibility of the reported 12.3% absolute IoU@0.4 improvement over OOAL in Table 2 and the split-robustness claims in Table 3. The authors must either demonstrate image-level disjointness for all splits of AGD20K-NAT and UMD-NAT, or rebuild the splits so that no image appears in more than one split, and re-report all comparisons.
  2. [Appendix A6, 'Detection-and-segmentation backbone'] The paper reports that Grounded-SAM detects an object in 281 of 345 test samples and that object-level recall is 80.29%. This means that for roughly one fifth of test samples the local branch cannot contribute and the method falls back to a uniform mask. The method's central premise is that the affordance query in Eq. (1) localizes the relevant object; when that premise fails, the reported test metrics are driven by a different mechanism. The paper should report results conditioned on detection success, and should check whether the claimed advantage over baselines holds on the subset of samples where detection succeeds, as well as on the fallback subset.
  3. [Sec. 5.3, Appendix A1] All main quantitative results are reported for a single random training-sample selection with no error bars or multiple seeds. On AGD20K-NAT the one-shot training set contains only 63 image-affordance pairs, so the sampling of the 50 egocentric images per object category could substantially change the outcome. The absence of variance estimates is especially problematic because the paper's headline conclusion is that AffordAnything+ 'consistently' outperforms baselines; without repeated trials or seeds, consistency is asserted, not demonstrated. Please report means and standard deviations over at least five random training-sample selections (and multiple seeds for the trainable modules) for Tables 2 and 3.
minor comments (6)
  1. [Eq. (8)] The notation f_t(f_I)^T / sqrt(d) is ambiguous: if f_t and f_I are both d-dimensional vectors, this expression is a scalar, not a spatial map. Please clarify how the global-branch prediction Y^G is formed from the image and text features, especially how a vector-level similarity becomes a pixel-level map.
  2. [Figure 2] The text encoder E_T is omitted from the diagram even though the text feature f_t is essential to the semantic fusion module; please add it or explicitly explain the omission in the caption.
  3. [Sec. 5.1, Appendix A1] The main text says 'the affordance types in the train, val, and test sets are mutually exclusive,' which is a statement about affordance types, not about images. Please state the actual split unit (image-affordance pair versus image) in the main text and make the appendix description consistent with it.
  4. [Table A12] In the table, the 'Ours' row claims quantitative evaluation of novel affordance types in 2D, but no existing 2D baseline in the table has a checkmark in the 'Aff.' column. The table is useful as a summary, but the novelty claim would be easier to evaluate if the quantitative NAT results for 2D baselines were referenced alongside the table.
  5. [Sec. A8] AffordAnything+ has a total inference time of 0.680 s versus 0.021 s for OOAL on the same GPU. The paper does not discuss the runtime-accuracy tradeoff in the main text; a brief sentence noting this cost and in which applications it is acceptable would help practitioners judge the method.
  6. [Sec. A2, 'AffordAnything'] The description says the default branch weight alpha is set to 0 for AffordAnything, which means the global branch is not used at all; the later ablation in Table A8 shows that initializing alpha to 0 during training for AffordAnything+ performs much worse than initializing to 0.5. This difference between the training-free default and the trainable initialization is worth explaining more explicitly in the main text.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: AffordAnything is an empirical pipeline with held-out test affordances; the only flagged concern is a possible validation/test image overlap, which is a benchmark-validity issue rather than circularity.

full rationale

The paper's central claims are empirical. AffordAnything uses the affordance label as a text query to Grounded-SAM for object detection and to CLIP for patch scoring, which is the task definition rather than a circular step. AffordAnything+ learns only a small fusion module on the NAT training split and selects the best checkpoint on the validation split; the test affordance types are disjoint from both, so no parameter is fitted to the test labels. The only passage resembling a concern is the pair-level split in Sec. A1 ('we decompose each image annotation into multiple image-affordance pairs'), which could in principle allow the same image to appear in both validation and test sets and thereby inflate the reported gains. However, this is a benchmark-validity issue, not a circular derivation: the reported IoU values are not equivalent by construction to any fitted input. The only self-citation ([38], by author Yeh) appears in related work on CLIP phrase localization and is not load-bearing for the NAT results. I find no step where a prediction is defined in terms of its own target or where a fitted constant is renamed as a prediction.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central result depends on pretrained components and on validation-set hyperparameters rather than on new postulated entities. The main epistemic cost is the reliance on Grounded-SAM's detection quality and on CLIP's ability to associate unseen action words with object parts; these are external capabilities, not derived in the paper.

free parameters (5)
  • number of evenly-focused patches e = 9
    Chosen via hyperparameter search on the AGD20K-NAT validation set (Tab. A7); performance is flat across 4, 9, 16.
  • border ratio b = 0.2
    Chosen via hyperparameter search on validation (Tab. A7).
  • branch weight init alpha = 0.5
    Initialization for the learned branch weight in AffordAnything+; selected on validation (Tab. A8). Initializing to 0 or 1 substantially changes performance.
  • learning rate = 1e-4 (AGD20K-NAT), 1e-5 (UMD-NAT)
    Selected by validation performance (Tab. A9).
  • default branch weight alpha for AffordAnything = 0, set to 1 when no object detected
    Hand-set in the training-free variant (Sec. A2), controlling the global/local branch blend.
assumptions (4)
  • domain assumption Affordance regions correlate with object subparts, so decomposing object masks into grid and border patches captures the relevant region.
    Motivates the patch refinement module (Fig. 3 and Sec. 4.2). If this correlation is weak, the local branch loses its advantage.
  • domain assumption Grounded-SAM, prompted only with the affordance label, detects the objects whose subregions are the affordance targets.
    Eq. (1) uses the affordance text as the detection query; detection recall is 80.29% on the test set (Sec. A6), so this premise fails on roughly one fifth of samples.
  • domain assumption Frozen CLIP image-text similarity on local masked patches ranks affordance-relevant subparts for novel affordance words.
    The local branch weights in Eq. (7) are softmax of cosine similarity between patch features and the affordance text feature; this assumes CLIP has enough compositional knowledge to match unseen action words to parts.
  • domain assumption The AGD20K and UMD annotations, re-split into disjoint affordance sets, provide a valid measure of generalization to novel affordance types.
    The benchmark contribution rests on the idea that held-out affordance words are truly novel rather than synonyms or near-duplicates of training words.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Zero-shot 2D Grounding with Novel Affordance Types." pith.science (2026). https://pith.science/paper/DI7RXLUZ

@misc{pith2026260808929,
  author       = {Pith},
  title        = {Pith review of: Zero-shot 2D Grounding with Novel Affordance Types},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DI7RXLUZ}},
  note         = {Machine review of arXiv:2608.08929}
}
read the original abstract

2D affordance grounding aims to locate the region of an object that a human can interact with. Existing research focuses on recognizing affordance types seen during training and does not study models' ability to generalize to novel affordances, which is crucial for real-world applications. We propose the task of zero-shot 2D grounding with novel affordance types (NAT) and introduce the NAT benchmarks. We then propose AffordAnything, a training-free method that leverages segmentation cues, motivated by the strong correlation between affordance regions and object subparts. To further improve performance, we develop AffordAnything+, a trainable variant that learns to combine these cues. On the proposed AGD20K-NAT benchmark, our best model AffordAnything+ achieves a substantial improvement of 12.3% (absolute) in IoU@0.4 over the SOTA affordance grounding method, OOAL.

Figures

Figures reproduced from arXiv: 2608.08929 by the authors.

Figure 1
Figure 1. We propose to study novel affordance types in the task of zero-shot 2D grounding. (a) Prior works in 2D zero-shot affordance grounding focus on novel object categories. (b) In contrast, we propose to address models’ generalization capability to novel affordance types. (c) The state-of￾the-art (SOTA) model OOAL [36] does not generalize to affordance types that are unseen during training, whereas our proposed method b… view at source ↗
Figure 2
Figure 2. Illustration of AffordAnything (Sec. 4.2). AffordAnything first generates local patch candidates through the (a) detection-and-segmentation backbone and the patch refinement module. Next, (b) the visual and text features are extracted and (c) fused in the semantic fusion module in a multi-branch manner. The frozen text encoder E T is omitted in the visualization for clarity. current approaches [36] still struggle to… view at source ↗
Figure 3
Figure 3. Motivation for local patches genera￾tion. The affordance region typically corresponds to a sub-region within the object mask. Given an image I ∈ R H×W×3 and an affor￾dance type label a expressed in natural language, e.g., ‘drink’, the task of 2D affordance ground￾ing aims at predicting the pixel-level affordance mask Y ∈ R H×W indicating the interactive region. To evaluate the model’s generalization ability to novel… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Patch refinement process. For each object mask, we generate evenly-focused patches ME n and border￾focused patches MB n to capture the local details based on the spatial relations. Generating local patches. Inspired by the observation that affordance regions strongly c…
Figure 5
Figure 5. Figure 5: Trainable fusion module in AffordAnything+ (Sec. 4.3). The global branch prediction Yˆ G and local branch predic￾tion Yˆ L are weighted by the learned branch weight α ∗ . Global branch decoder. Given the image fea￾ture f I and text feature f t , the global branch de￾co…
Figure 6
Figure 6. Figure 6: Qualitative results on AGD20K-NAT test set. We compare our AffordAnything+ against other affordance grounding and fine-tuned open-vocabulary semantic segmentation baselines. Our method consistently localizes affordance-relevant subregions for novel affordance types. sh…
Figure 7
Figure 7. Figure 7: Additional qualitative results on generalization. We show the predictions on (a)(b) base object categories but novel affordances; (c) novel object categories but base affordances; and (d) novel object categories and novel affordances [PITH_FULL_IMAGE:figures/full_fig_…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

90 extracted references · 70 canonical work pages

  1. [1]

    GPT-4 technical report.arXiv preprint arXiv:2303.08774, 2023

    Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, et al. GPT-4 technical report.arXiv preprint arXiv:2303.08774, 2023. 2

    arXiv 2023

  2. [2]

    Do as I can, not as I say: Grounding language in robotic affordances

    Michael Ahn, Anthony Brohan, Noah Brown, Yevgen Chebotar, Omar Cortes, Byron David, Chelsea Finn, Chuyuan Fu, Keerthana Gopalakrishnan, Karol Hausman, et al. Do as I can, not as I say: Grounding language in robotic affordances. InCORL, 2022. 1

    2022

  3. [3]

    Affordances from human videos as a versatile representation for robotics

    Shikhar Bahl, Russell Mendonca, Lili Chen, Unnat Jain, and Deepak Pathak. Affordances from human videos as a versatile representation for robotics. InCVPR, 2023. 1

    2023

  4. [4]

    ViSPLA: Visual iterative self-prompting for language-guided 3d affordance learning

    Hritam Basak and Zhaozheng Yin. ViSPLA: Visual iterative self-prompting for language-guided 3d affordance learning. InNeurIPS, 2025. 9

    2025

  5. [5]

    Grounding everything: Emerging localization properties in vision-language transformers

    Walid Bousselham, Felix Petersen, Vittorio Ferrari, and Hilde Kuehne. Grounding everything: Emerging localization properties in vision-language transformers. InCVPR, 2024. 2, 6, 7, 3

    2024

  6. [6]

    Language models are few-shot learners

    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. InNeurIPS, 2020. 2

    2020

  7. [7]

    Emerging properties in self-supervised vision transformers

    Mathilde Caron, Hugo Touvron, Ishan Misra, Hervé Jégou, Julien Mairal, Piotr Bojanowski, and Armand Joulin. Emerging properties in self-supervised vision transformers. InICCV, 2021. 2

    2021

  8. [8]

    Worldafford: Affordance grounding based on natural language instructions

    Changmao Chen, Yuren Cong, and Zhen Kan. Worldafford: Affordance grounding based on natural language instructions. InICTAI, 2024. 2, 6, 8

    2024

Show all 90 references
  1. [9]

    Affordance grounding from demon- stration video to target image

    Joya Chen, Difei Gao, Kevin Qinghong Lin, and Mike Zheng Shou. Affordance grounding from demon- stration video to target image. InCVPR, 2023. 2

    2023

  2. [10]

    Affordances of augmented reality in science learning: Suggestions for future research.JSET, 2013

    Kun-Hung Cheng and Chin-Chung Tsai. Affordances of augmented reality in science learning: Suggestions for future research.JSET, 2013. 1

    2013

  3. [11]

    3D- AffordanceLLM: Harnessing large language models for open-vocabulary affordance detection in 3D worlds

    Hengshuo Chu, Xiang Deng, Xiaoyang Chen, Yinchuan Li, Jianye Hao, and Liqiang Nie. 3D- AffordanceLLM: Harnessing large language models for open-vocabulary affordance detection in 3D worlds. InICLR, 2025. 2, 9

    2025

  4. [12]

    Learning to act properly: Predicting and explaining affordances from images

    Ching-Yao Chuang, Jiaman Li, Antonio Torralba, and Sanja Fidler. Learning to act properly: Predicting and explaining affordances from images. InCVPR, 2018. 2

    2018

  5. [13]

    A deep multi-level network for saliency prediction

    Marcella Cornia, Lorenzo Baraldi, Giuseppe Serra, and Rita Cucchiara. A deep multi-level network for saliency prediction. InICPR, 2016. 2

    2016

  6. [14]

    Strap: Structured object affordance segmentation with point supervision.arXiv preprint arXiv:2304.08492, 2023

    Leiyao Cui, Xiaoxue Chen, Hao Zhao, Guyue Zhou, and Yixin Zhu. Strap: Structured object affordance segmentation with point supervision.arXiv preprint arXiv:2304.08492, 2023. 2

    2023 arXiv

  7. [15]

    What does CLIP know about peeling a banana? InCVPRW, 2024

    Claudia Cuttano, Gabriele Rosi, Gabriele Trivigno, and Giuseppe Averta. What does CLIP know about peeling a banana? InCVPRW, 2024. 2, 9

    2024

  8. [16]

    Decoupling zero-shot semantic segmentation

    Jian Ding, Nan Xue, Gui-Song Xia, and Dengxin Dai. Decoupling zero-shot semantic segmentation. In CVPR, 2022. 3

    2022

  9. [17]

    Affordancenet: An end-to-end deep learning approach for object affordance detection

    Thanh-Toan Do, Anh Nguyen, and Ian Reid. Affordancenet: An end-to-end deep learning approach for object affordance detection. InICRA, 2018. 2

    2018

  10. [18]

    Demo2vec: Reasoning object affordances from online videos

    Kuan Fang, Te-Lin Wu, Daniel Yang, Silvio Savarese, and Joseph J Lim. Demo2vec: Reasoning object affordances from online videos. InCVPR, 2018. 2

    2018

  11. [19]

    Learning 2d invariant affordance knowledge for 3d affordance grounding

    Xianqiang Gao, Pingrui Zhang, Delin Qu, Dong Wang, Zhigang Wang, Yan Ding, and Bin Zhao. Learning 2d invariant affordance knowledge for 3d affordance grounding. InAAAI, 2025. 9

    2025

  12. [20]

    RLAfford: End-to-end affordance learning for robotic manipulation

    Yiran Geng, Boshi An, Haoran Geng, Yuanpei Chen, Yaodong Yang, and Hao Dong. RLAfford: End-to-end affordance learning for robotic manipulation. InICRA, 2023. 1

    2023

  13. [21]

    Psychology press, 2014

    James J Gibson.The ecological approach to visual perception: classic edition. Psychology press, 2014. 1, 8 10

    2014

  14. [22]

    One-shot transfer of affordance regions? AffCorrs! InCORL, 2023

    Denis Hadjivelichkov, Sicelukwanda Zwane, Lourdes Agapito, Marc Peter Deisenroth, and Dimitrios Kanoulas. One-shot transfer of affordance regions? AffCorrs! InCORL, 2023. 2

    2023

  15. [23]

    Visual affordance and function understanding: A survey.ACM Comput

    Mohammed Hassanin, Salman Khan, and Murat Tahtali. Visual affordance and function understanding: A survey.ACM Comput. Surv., 2021. 1

    2021

  16. [24]

    V oxPoser: Composable 3d value maps for robotic manipulation with language models

    Wenlong Huang, Chen Wang, Ruohan Zhang, Yunzhu Li, Jiajun Wu, and Li Fei-Fei. V oxPoser: Composable 3d value maps for robotic manipulation with language models. InCORL, 2023. 2

    2023

  17. [25]

    Predicting gaze in egocentric video by learning task-dependent attention transition

    Yifei Huang, Minjie Cai, Zhenqiang Li, and Yoichi Sato. Predicting gaze in egocentric video by learning task-dependent attention transition. InECCV, 2018. 2

    2018

  18. [26]

    Resource-efficient affordance grounding with complementary depth and semantic prompts

    Yizhou Huang, Fan Yang, Guoliang Zhu, Gen Li, Hao Shi, Yukun Zuo, Wenrui Chen, Zhiyong Li, and Kailun Yang. Resource-efficient affordance grounding with complementary depth and semantic prompts. InIROS, 2025. 2

    2025

  19. [27]

    Intra: Interaction relationship-aware weakly supervised affordance grounding

    Ji Ha Jang, Hoigi Seo, and Se Young Chun. Intra: Interaction relationship-aware weakly supervised affordance grounding. InECCV, 2024. 2, 9

    2024

  20. [28]

    One-shot affordance grounding of deformable objects in egocentric organizing scenes

    Wanjun Jia, Fan Yang, Mengfei Duan, Xianchi Chen, Yinxi Wang, Yiming Jiang, Wenrui Chen, Kailun Yang, and Zhiyong Li. One-shot affordance grounding of deformable objects in egocentric organizing scenes. InIROS, 2025. 2

    2025

  21. [29]

    AffordanceSAM: Segment anything once more in affordance grounding.arXiv preprint arXiv:2504.15650, 2025

    Dengyang Jiang, Zanyi Wang, Hengzhuang Li, Sizhe Dang, Teli Ma, Wei Wei, Guang Dai, Lei Zhang, and Mengmeng Wang. AffordanceSAM: Segment anything once more in affordance grounding.arXiv preprint arXiv:2504.15650, 2025. 2

    2025 arXiv

  22. [30]

    Robo-ABC: Affordance generalization beyond categories via semantic correspondence for robot manipulation

    Yuanchen Ju, Kaizhe Hu, Guowei Zhang, Gu Zhang, Mingrun Jiang, and Huazhe Xu. Robo-ABC: Affordance generalization beyond categories via semantic correspondence for robot manipulation. In ECCV, 2024. 2, 9

    2024

  23. [31]

    Learning human activities and object affordances from RGB-D videos.IJRR, 2013

    Hema Swetha Koppula, Rudhir Gupta, and Ashutosh Saxena. Learning human activities and object affordances from RGB-D videos.IJRR, 2013. 2

    2013

  24. [32]

    Deepgaze ii: Reading fixations from deep features trained on object recognition.arXiv preprint arXiv:1610.01563, 2016

    Matthias Kümmerer, Thomas SA Wallis, and Matthias Bethge. Deepgaze ii: Reading fixations from deep features trained on object recognition.arXiv preprint arXiv:1610.01563, 2016. 2

    2016 arXiv

  25. [33]

    ClearCLIP: Decomposing CLIP representations for dense vision-language inference

    Mengcheng Lan, Chaofeng Chen, Yiping Ke, Xinjiang Wang, Litong Feng, and Wayne Zhang. ClearCLIP: Decomposing CLIP representations for dense vision-language inference. InECCV, 2024. 2, 3, 6, 7

    2024

  26. [34]

    ProxyCLIP: Proxy attention improves CLIP for open-vocabulary segmentation

    Mengcheng Lan, Chaofeng Chen, Yiping Ke, Xinjiang Wang, Litong Feng, and Wayne Zhang. ProxyCLIP: Proxy attention improves CLIP for open-vocabulary segmentation. InECCV, 2024. 2, 3, 6, 7

    2024

  27. [35]

    Locate: Localize and transfer object parts for weakly supervised affordance grounding

    Gen Li, Varun Jampani, Deqing Sun, and Laura Sevilla-Lara. Locate: Localize and transfer object parts for weakly supervised affordance grounding. InCVPR, 2023. 2, 8, 9

    2023

  28. [36]

    One-shot open affordance learning with foundation models

    Gen Li, Deqing Sun, Laura Sevilla-Lara, and Varun Jampani. One-shot open affordance learning with foundation models. InCVPR, 2024. 1, 2, 3, 6, 7, 8, 4, 5, 9

    2024

  29. [37]

    Learning precise affordances from egocentric videos for robotic manipulation

    Gen Li, Nikolaos Tsagkas, Jifei Song, Ruaridh Mon-Williams, Sethu Vijayakumar, Kun Shao, and Laura Sevilla-Lara. Learning precise affordances from egocentric videos for robotic manipulation. InICCV,

  30. [38]

    Adapting CLIP for phrase localization without further training.arXiv preprint arXiv:2204.03647, 2022

    Jiahao Li, Greg Shakhnarovich, and Raymond A Yeh. Adapting CLIP for phrase localization without further training.arXiv preprint arXiv:2204.03647, 2022. 3

    2022 arXiv

  31. [39]

    Laso: Language-guided affordance segmentation on 3d object

    Yicong Li, Na Zhao, Junbin Xiao, Chun Feng, Xiang Wang, and Tat-seng Chua. Laso: Language-guided affordance segmentation on 3d object. InCVPR, 2024. 9

    2024

  32. [40]

    Intermediate connectors and geometric priors for language-guided affordance segmentation on unseen object categories

    Yicong Li, Yiyang Chen, Zhenyuan Ma, Junbin Xiao, Xiang Wang, and Angela Yao. Intermediate connectors and geometric priors for language-guided affordance segmentation on unseen object categories. InICCV, 2025. 9

    2025

  33. [41]

    Grounding DINO: Marrying dino with grounded pre-training for open-set object detection

    Shilong Liu, Zhaoyang Zeng, Tianhe Ren, Feng Li, Hao Zhang, Jie Yang, Qing Jiang, Chunyuan Li, Jianwei Yang, Hang Su, et al. Grounding DINO: Marrying dino with grounded pre-training for open-set object detection. InECCV, 2024. 2

    2024

  34. [42]

    GEAL: Generalizable 3d affordance learning with cross-modal consistency

    Dongyue Lu, Lingdong Kong, Tianxin Huang, and Gim Hee Lee. GEAL: Generalizable 3d affordance learning with cross-modal consistency. InCVPR, 2025. 2, 9 11

    2025

  35. [43]

    Learning affordance grounding from exocentric images

    Hongchen Luo, Wei Zhai, Jing Zhang, Yang Cao, and Dacheng Tao. Learning affordance grounding from exocentric images. InCVPR, 2022. 1, 2, 3, 6, 8, 4, 9

    2022

  36. [44]

    Learning visual affordance grounding from demonstration videos.IEEE TNNLS, 2023

    Hongchen Luo, Wei Zhai, Jing Zhang, Yang Cao, and Dacheng Tao. Learning visual affordance grounding from demonstration videos.IEEE TNNLS, 2023. 2

    2023

  37. [45]

    Grounding language with visual affordances over unstructured data

    Oier Mees, Jessica Borja-Diaz, and Wolfram Burgard. Grounding language with visual affordances over unstructured data. InICRA, 2023. 2

    2023

  38. [46]

    LAN-grasp: Using large language models for semantic object grasping.arXiv preprint arXiv:2310.05239, 2023

    Reihaneh Mirjalili, Michael Krawez, Simone Silenzi, Yannik Blei, and Wolfram Burgard. LAN-grasp: Using large language models for semantic object grasping.arXiv preprint arXiv:2310.05239, 2023. 2

    2023

  39. [47]

    Embodied large language models enable robots to complete complex tasks in unpredictable environments.Nature Machine Intelligence,

    Ruaridh Mon-Williams, Gen Li, Ran Long, Wenqian Du, and Chris Lucas. Embodied large language models enable robots to complete complex tasks in unpredictable environments.Nature Machine Intelligence,

  40. [48]

    Selective contrastive learning for weakly supervised affordance grounding

    WonJun Moon, Hyun Seok Seong, and Jae-Pil Heo. Selective contrastive learning for weakly supervised affordance grounding. InICCV, 2025. 2, 9

    2025

  41. [49]

    Affordance detection of tool parts from geometric features

    Austin Myers, Ching L Teo, Cornelia Fermüller, and Yiannis Aloimonos. Affordance detection of tool parts from geometric features. InICRA, 2015. 2

    2015

  42. [50]

    Learning affordance landscapes for interaction exploration in 3d environments

    Tushar Nagarajan and Kristen Grauman. Learning affordance landscapes for interaction exploration in 3d environments. InNeurIPS, 2020. 2

    2020

  43. [51]

    Grounded human-object interaction hotspots from video

    Tushar Nagarajan, Christoph Feichtenhofer, and Kristen Grauman. Grounded human-object interaction hotspots from video. InICCV, 2019. 2

    2019

  44. [52]

    Caldwell, and Nikos G

    Anh Nguyen, Dimitrios Kanoulas, Darwin G. Caldwell, and Nikos G. Tsagarakis. Object-based affordances detection with convolutional neural networks and dense conditional random fields. InIROS, 2017. 1, 2, 3, 6

    2017

  45. [53]

    Open- vocabulary affordance detection in 3d point clouds

    Toan Nguyen, Minh Nhat Vu, An Vuong, Dzung Nguyen, Thieu V o, Ngan Le, and Anh Nguyen. Open- vocabulary affordance detection in 3d point clouds. InIROS, 2023. 2, 3, 9

    2023

  46. [54]

    Introducing GPT-5.2

    OpenAI. Introducing GPT-5.2. https://openai.com/index/introducing-gpt-5-2/ , 2026. Ac- cessed: 2026-02-23. 6, 3

    2026

  47. [55]

    Dinov2: Learning robust visual features without supervision.TMLR, 2024

    Maxime Oquab, Timothée Darcet, Théo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al. Dinov2: Learning robust visual features without supervision.TMLR, 2024. 2, 6

    2024

  48. [56]

    SalGAN: Visual saliency prediction with generative adversarial networks.arXiv preprint arXiv:1701.01081, 2017

    Junting Pan, Cristian Canton Ferrer, Kevin McGuinness, Noel E O’Connor, Jordi Torres, Elisa Sayrol, and Xavier Giro-i Nieto. SalGAN: Visual saliency prediction with generative adversarial networks.arXiv preprint arXiv:1701.01081, 2017. 2

    2017 arXiv

  49. [57]

    AffordanceLLM: Grounding affordance from vision language models

    Shengyi Qian, Weifeng Chen, Min Bai, Xiong Zhou, Zhuowen Tu, and Li Erran Li. AffordanceLLM: Grounding affordance from vision language models. InCVPRW, 2024. 2, 6, 9

    2024

  50. [58]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. InICML, 2021. 4, 6, 2

    2021

  51. [59]

    Strategies to leverage foundational model knowledge in object affordance grounding

    Arushi Rai, Kyle Buettner, and Adriana Kovashka. Strategies to leverage foundational model knowledge in object affordance grounding. InCVPRW, 2024. 2, 9

    2024

  52. [60]

    Language embedded radiance fields for zero-shot task-oriented grasping

    Adam Rashid, Satvik Sharma, Chung Min Kim, Justin Kerr, Lawrence Yunliang Chen, Angjoo Kanazawa, and Ken Goldberg. Language embedded radiance fields for zero-shot task-oriented grasping. InCoRL,

  53. [61]

    SAM2: Segment anything in images and videos.arXiv preprint arXiv:2408.00714, 2024

    Nikhila Ravi, Valentin Gabeur, Yuan-Ting Hu, Ronghang Hu, Chaitanya Ryali, Tengyu Ma, Haitham Khedr, Roman Rädle, Chloe Rolland, Laura Gustafson, et al. SAM2: Segment anything in images and videos.arXiv preprint arXiv:2408.00714, 2024. 2

    2024 arXiv

  54. [62]

    Grounded SAM: Assembling open-world models for diverse visual tasks.arXiv preprint arXiv:2401.14159, 2024

    Tianhe Ren, Shilong Liu, Ailing Zeng, Jing Lin, Kunchang Li, He Cao, Jiayu Chen, Xinyu Huang, Yukang Chen, Feng Yan, et al. Grounded SAM: Assembling open-world models for diverse visual tasks.arXiv preprint arXiv:2401.14159, 2024. 4, 9, 2, 5

    2024 arXiv

  55. [63]

    Adaptive binarization for weakly supervised affordance segmentation

    Johann Sawatzky and Jurgen Gall. Adaptive binarization for weakly supervised affordance segmentation. InICCVW, 2017. 2 12

    2017

  56. [64]

    Weakly supervised affordance detection

    Johann Sawatzky, Abhilash Srikantha, and Juergen Gall. Weakly supervised affordance detection. In CVPR, 2017. 2

    2017

  57. [65]

    GREAT: Geometry- intention collaborative inference for open-vocabulary 3d object affordance grounding

    Yawen Shao, Wei Zhai, Yuhang Yang, Hongchen Luo, Yang Cao, and Zheng-Jun Zha. GREAT: Geometry- intention collaborative inference for open-vocabulary 3d object affordance grounding. InCVPR, 2025. 3, 9

    2025

  58. [66]

    Learning 6-DoF fine-grained grasp detection based on part affordance grounding.IEEE TASE, 2025

    Yaoxian Song, Penglei Sun, Piaopiao Jin, Yi Ren, Yu Zheng, Zhixu Li, Xiaowen Chu, Yue Zhang, Tiefeng Li, and Jason Gu. Learning 6-DoF fine-grained grasp detection based on part affordance grounding.IEEE TASE, 2025. 2

    2025

  59. [67]

    Closed-loop transfer for weakly-supervised affordance grounding

    Jiajin Tang, Zhengxuan Wei, Ge Zheng, and Sibei Yang. Closed-loop transfer for weakly-supervised affordance grounding. InICCV, 2025. 2, 9

    2025

  60. [68]

    Oval-prompt: Open-vocabulary affordance localization for robot manipulation through llm affordance-grounding

    Edmond Tong, Anthony Opipari, Stanley Lewis, Zhen Zeng, and Odest Chadwicke Jenkins. Oval-prompt: Open-vocabulary affordance localization for robot manipulation through llm affordance-grounding. In ICRAW, 2024. 2, 9

    2024

  61. [69]

    Open- vocabulary affordance detection using knowledge distillation and text-point correlation

    Tuan Van V o, Minh Nhat Vu, Baoru Huang, Toan Nguyen, Ngan Le, Thieu V o, and Anh Nguyen. Open- vocabulary affordance detection using knowledge distillation and text-point correlation. InICRA, 2024. 3

    2024

  62. [70]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. InNeurIPS, 2017. 6

    2017

  63. [71]

    Predicting actions from static scenes

    Tuan-Hung Vu, Catherine Olsson, Ivan Laptev, Aude Oliva, and Josef Sivic. Predicting actions from static scenes. InECCV, 2014. 1

    2014

  64. [72]

    SCLIP: Rethinking self-attention for dense vision-language inference

    Feng Wang, Jieru Mei, and Alan Yuille. SCLIP: Rethinking self-attention for dense vision-language inference. InECCV, 2024. 2, 6, 7, 3

    2024

  65. [73]

    AffordDexGrasp: Open-set language-guided dexterous grasp with generalizable-instructive affordance

    Yi-Lin Wei, Mu Lin, Yuhao Lin, Jian-Jian Jiang, Xiao-Ming Wu, Ling-An Zeng, and Wei-Shi Zheng. AffordDexGrasp: Open-set language-guided dexterous grasp with generalizable-instructive affordance. In ICCV, 2025. 3

    2025

  66. [74]

    3DAffordSplat: Efficient affordance reasoning with 3d gaussians

    Zeming Wei, Junyi Lin, Yang Liu, Weixing Chen, Jingzhou Luo, Guanbin Li, and Liang Lin. 3DAffordSplat: Efficient affordance reasoning with 3d gaussians. InACMMM, 2025. 2, 9

    2025

  67. [75]

    RAGNet: Large-scale reasoning-based affordance segmentation benchmark towards general grasping

    Dongming Wu, Yanping Fu, Saike Huang, Yingfei Liu, Fan Jia, Nian Liu, Feng Dai, Tiancai Wang, Rao Muhammad Anwer, Fahad Shahbaz Khan, et al. RAGNet: Large-scale reasoning-based affordance segmentation benchmark towards general grasping. InICCV, 2025. 2

    2025

  68. [76]

    Towards open vocabulary learning: A survey.IEEE TPAMI, 2024

    Jianzong Wu, Xiangtai Li, Shilin Xu, Haobo Yuan, Henghui Ding, Yibo Yang, Xia Li, Jiangning Zhang, Yunhai Tong, Xudong Jiang, et al. Towards open vocabulary learning: A survey.IEEE TPAMI, 2024. 3

    2024

  69. [77]

    Florence-2: Advancing a unified representation for a variety of vision tasks

    Bin Xiao, Haiping Wu, Weijian Xu, Xiyang Dai, Houdong Hu, Yumao Lu, Michael Zeng, Ce Liu, and Lu Yuan. Florence-2: Advancing a unified representation for a variety of vision tasks. InCVPR, 2024. 6, 7

    2024

  70. [78]

    SED: A simple encoder-decoder for open-vocabulary semantic segmentation

    Bin Xie, Jiale Cao, Jin Xie, Fahad Shahbaz Khan, and Yanwei Pang. SED: A simple encoder-decoder for open-vocabulary semantic segmentation. InCVPR, 2024. 3

    2024

  71. [79]

    Weakly supervised multimodal affordance grounding for egocentric images

    Lingjing Xu, Yang Gao, Wenfeng Song, and Aimin Hao. Weakly supervised multimodal affordance grounding for egocentric images. InAAAI, 2024. 2, 8, 9

    2024

  72. [80]

    Side adapter network for open- vocabulary semantic segmentation

    Mengde Xu, Zheng Zhang, Fangyun Wei, Han Hu, and Xiang Bai. Side adapter network for open- vocabulary semantic segmentation. InCVPR, 2023. 3

    2023

  73. [81]

    Weakly-supervised affordance grounding guided by part-level semantic priors

    Peiran Xu and MU Yadong. Weakly-supervised affordance grounding guided by part-level semantic priors. InICLR, 2025. 2, 9

    2025

  74. [82]

    Universal instance perception as object discovery and retrieval

    Bin Yan, Yi Jiang, Jiannan Wu, Dong Wang, Ping Luo, Zehuan Yuan, and Huchuan Lu. Universal instance perception as object discovery and retrieval. InCVPR, 2023. 5

    2023

  75. [83]

    Grounding 3d object affordance from 2d interactions in images

    Yuhang Yang, Wei Zhai, Hongchen Luo, Yang Cao, Jiebo Luo, and Zheng-Jun Zha. Grounding 3d object affordance from 2d interactions in images. InICCV, 2023. 9

    2023

  76. [84]

    2HandedAf- forder: Learning precise actionable bimanual affordances from human videos

    Chunlin Yu, Hanqing Wang, Ye Shi, Haoyang Luo, Sibei Yang, Jingyi Yu, and Jingya Wang. 2HandedAf- forder: Learning precise actionable bimanual affordances from human videos. InICCV, 2025. 2 13

    2025

  77. [85]

    SeqAfford: Sequential 3d affordance reasoning via multimodal large language model

    Chunlin Yu, Hanqing Wang, Ye Shi, Haoyang Luo, Sibei Yang, Jingyi Yu, and Jingya Wang. SeqAfford: Sequential 3d affordance reasoning via multimodal large language model. InCVPR, 2025. 2, 9

    2025

  78. [86]

    One-shot object affordance detection in the wild.IJCV, 2022

    Wei Zhai, Hongchen Luo, Jing Zhang, Yang Cao, and Dacheng Tao. One-shot object affordance detection in the wild.IJCV, 2022. 2, 9

    2022

  79. [87]

    Extract free dense labels from CLIP

    Chong Zhou, Chen Change Loy, and Bo Dai. Extract free dense labels from CLIP. InECCV, 2022. 3

    2022

  80. [88]

    ZegClip: Towards adapting CLIP for zero-shot semantic segmentation

    Ziqin Zhou, Yinjie Lei, Bowen Zhang, Lingqiao Liu, and Yifan Liu. ZegClip: Towards adapting CLIP for zero-shot semantic segmentation. InCVPR, 2023. 3

    2023

  81. [89]

    Grounding 3d object affordance with language instructions, visual observations and interactions

    He Zhu, Quyu Kong, Kechun Xu, Xunlong Xia, Bing Deng, Jieping Ye, Rong Xiong, and Yue Wang. Grounding 3d object affordance with language instructions, visual observations and interactions. InCVPR,

  82. [2025]

    Seen” and “Unseen

    9 14 Appendix The appendix is organized as follows: • In Sec. A1, we provide details on the datasets and splits for the proposed NAT benchmarks. • In Sec. A2, we provide implementation details on our proposed models. • In Sec. A3, we provide implementation details on the basel...

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.