pith. sign in

arxiv: 2606.27307 · v1 · pith:IJ3SFR4Bnew · submitted 2026-06-25 · 💻 cs.CV

See & Sniff: Learning Visuo-Olfactory Representations

Seongyu Kim , Seungwoo Lee , Hyeonggon Ryu , Joon Son Chung , Arda Senocak This is my paper

Pith reviewed 2026-06-26 05:28 UTC · model grok-4.3

classification 💻 cs.CV
keywords visuo-olfactory representationsself-supervised learningmultimodal datasetsmell localizationcross-modal retrievalodor classificationsaliency mapssynthetic pairing
0
0 comments X

The pith

Synthetically pairing smell samples with semantically matched web images enables self-supervised visuo-olfactory representations that raise smell classification accuracy by 7%.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper shows that odor identity stays consistent within semantic categories, so existing smell-only data can be matched with in-the-wild web images to create a large paired visuo-olfactory dataset called SmellNet-V without new collection. On this dataset the See & Sniff framework applies self-supervised dense local alignment between visual patches and smell features to produce joint representations. These representations improve smell-only classification by 7 percent, support cross-modal retrieval, and generate saliency maps that ground odors spatially for a new localization task. A sympathetic reader cares because olfaction has remained absent from most multimodal models due to paired-data scarcity, and the method offers a scalable workaround.

Core claim

Odor identity is largely invariant to visual transformations within a semantic category. This invariance permits synthetic pairing of smell-only samples with aligned web images to form SmellNet-V. See & Sniff then learns joint visuo-olfactory representations through self-supervised dense local alignment, which improves smell classification from smell input alone by 7 percent over baselines, enables cross-modal retrieval, and produces smell saliency maps for pixel-level smell localization.

What carries the argument

See & Sniff self-supervised framework that performs dense local alignment between visual and olfactory features on the synthetically paired SmellNet-V dataset, yielding joint representations and smell saliency maps.

If this is right

  • Smell classification accuracy from olfactory input alone rises by 7 percent relative to smell-only baselines.
  • The learned representations support retrieval between images and smell samples across modalities.
  • Saliency maps from the model enable evaluation on a new pixel-level smell localization benchmark.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same semantic-invariance pairing trick could bootstrap multimodal datasets for other data-scarce senses such as taste or touch.
  • Spatial smell grounding via saliency maps may transfer to embodied systems that must locate odor sources in visual scenes.

Load-bearing premise

Odor identity remains the same even when the visual appearance of the source object changes within the same semantic category.

What would settle it

Collect a set of truly co-located vision and smell recordings, train See & Sniff on the synthetic pairs, and test whether the model still outperforms smell-only baselines on classification or localization; equal or worse performance would falsify the value of the synthetic pairing.

Figures

Figures reproduced from arXiv: 2606.27307 by Arda Senocak, Hyeonggon Ryu, Joon Son Chung, Seongyu Kim, Seungwoo Lee.

Figure 1
Figure 1. Figure 1: What can See & Sniff do? We show that See & Sniff, the framework that learns joint visuo-olfactory representations, can handle both unimodal and cross-modal tasks. (Left) Smell localization identifies the locations of smell sources within a visual scene based on input olfactory signals. (Right Top) Cross-modal retrieval demonstrates semantic alignment through bidirectional retrieval. (Right Bottom) Smell c… view at source ↗
Figure 2
Figure 2. Figure 2: Pipeline of See & Sniff. Our framework expands smell-only data through semantic pairing with web images. Vision and smell encoders extract modality-specific features, which are aligned using a contrastive objective to learn joint visuo-olfactory representations. 3.2 Model and Training Objective Contrastive Learning aims to learn aligned representations by attracting pos￾itive pairs while repelling negative… view at source ↗
Figure 3
Figure 3. Figure 3: Family-wise Comparison. 4.3 Main Results – Smell Classification Task We evaluate the quality of learned olfactory representations on ingredient classifi￾cation across the 50 categories defined in SmellNet. After self-supervised training of See & Sniff on SmellNet-V, we freeze the olfactory encoder and train a linear probe using unimodal olfactory signals with the labels from SmellNet. Evalua￾tion is conduc… view at source ↗
Figure 4
Figure 4. Figure 4: Smell→Vision Retrieval. The top–10 images retrieved by given smell queries are shown. Blue borders indicate correct matches, and red borders indicate mismatches. The failure cases are visually similar to the queried ingredients, indicating fine-grained visual ambiguity rather than random mismatch. sual supervision. Nevertheless, the consistent improvement across all ingredient families suggests that visuo-… view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative Smell Localization Results. Qualitative Results. The results are in [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Qualitative Results on Interactive Localization. [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Localization across Physical States. In contrast, See & Sniff outperforms all base￾lines except the upper-bound setting. These find￾ings indicate that interactive smell grounding re￾quires fine-grained, smell-conditioned spatial rea￾soning rather than static saliency or global embed￾ding similarity. Although the overall IoU values reflect the difficulty of the task, our model con￾sistently shows superior c… view at source ↗
Figure 8
Figure 8. Figure 8: Prompt for Search Query Generation. query refinement, automated image crawling, and the systematic removal of corrupted or duplicate samples to ensure the integrity of the final collection. Image Filtering. To ensure semantic precision and visual high-fidelity, we im￾plement a rigorous three-stage filtering pipeline: (1) prompt-based verification, (2) quality filtering, and (3) human refinement. 1. Categor… view at source ↗
Figure 9
Figure 9. Figure 9: Positive and Neg￾ative prompt pairs for CLIP verification. First, we perform prompt-based verification us￾ing CLIP [36] to validate candidates across three criteria: category consistency, photorealism, and object state. In particular, the state-based filter excludes degraded or spoiled instances, ensuring the data aligns with the standard, non-degraded states of ingredients typically sourced from public re… view at source ↗
Figure 10
Figure 10. Figure 10: Custom-built Gradio Pages for Image Collection (Top) and Image [PITH_FULL_IMAGE:figures/full_fig_p020_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Zero-shot Localization. 14 Additional Qualitative Results Due to space constraints in the main paper, only a selected subset of qualitative results was presented. In this supplementary material, we provide comprehensive visualizations to further demonstrate the robustness of our model [PITH_FULL_IMAGE:figures/full_fig_p027_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Qualitative Smell Localization Results [PITH_FULL_IMAGE:figures/full_fig_p028_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Qualitative Results on Interactive Localization. [PITH_FULL_IMAGE:figures/full_fig_p029_13.png] view at source ↗
read the original abstract

While modern multimodal models integrate vision with language, audio, or touch, olfaction remains largely unexplored due to the lack of paired visuo-olfactory data. We introduce SmellNet-V, a scalable visuo-olfactory dataset built on the insight that odor identity is largely invariant to visual transformations within a semantic category. This allows us to synthetically pair smell-only samples with semantically aligned in-the-wild web images, converting a unimodal olfactory dataset into a cross-modal benchmark without costly co-collection. Building on this dataset, we propose See & Sniff, a self-supervised framework that learns joint visuo-olfactory representations via dense local alignment and naturally produces smell saliency maps for spatial grounding of odor sources. We further introduce pixel-level smell localization task and a benchmark for evaluation. Our method surpasses smell-only baselines by 7% in smell classification from smell alone and generalizes to cross-modal retrieval and smell localization, establishing visuo-olfactory learning as a new direction in multimodal perception.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 1 minor

Summary. The manuscript introduces SmellNet-V, a visuo-olfactory dataset constructed by synthetically pairing smell-only samples with semantically matched in-the-wild web images under the assumption that odor identity is largely invariant to visual transformations within a semantic category. It proposes the See & Sniff self-supervised framework that learns joint representations through dense local alignment and generates smell saliency maps for spatial grounding. The method reports a 7% improvement over smell-only baselines in smell classification and shows generalization to cross-modal retrieval and a new pixel-level smell localization task and benchmark.

Significance. If the synthetic pairing holds and the gains are reproducible, the work offers a scalable route to multimodal olfaction research without co-collection costs and introduces a useful new localization benchmark. The self-supervised dense alignment approach is a sensible technical direction for this modality pair.

major comments (3)
  1. [§3 (Dataset Construction)] The invariance assumption ('odor identity is largely invariant to visual transformations within a semantic category') used to justify synthetic pairing of smell samples with web images is stated without any reported validation (e.g., human agreement scores on pairing quality, odor consistency checks across visual instances, or ablation versus real co-collected pairs). This assumption is load-bearing for the SmellNet-V dataset quality, the dense local alignment objective, and the derived saliency maps.
  2. [§5 (Experiments)] The claimed 7% improvement in smell classification from smell alone is presented without error bars, statistical significance tests, dataset statistics, or ablations that isolate the contribution of the visual pairing versus the self-supervised objective. This makes it impossible to determine whether the lift is robust or an artifact of the synthetic construction.
  3. [§4 (Method)] The dense local alignment loss and the procedure for extracting smell saliency maps from the learned representations are described only at a high level; no explicit equations or algorithmic details are provided for either component, preventing assessment of their technical novelty relative to prior cross-modal alignment methods.
minor comments (1)
  1. [Abstract] Quantitative results for cross-modal retrieval and smell localization are mentioned in the abstract but not reported with numbers or tables in the provided text, weakening the generalization claim.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their thoughtful comments, which highlight important areas for improvement in our manuscript. We address each major comment below and commit to revisions that strengthen the work.

read point-by-point responses

  1. Referee: [§3 (Dataset Construction)] The invariance assumption ('odor identity is largely invariant to visual transformations within a semantic category') used to justify synthetic pairing of smell samples with web images is stated without any reported validation (e.g., human agreement scores on pairing quality, odor consistency checks across visual instances, or ablation versus real co-collected pairs). This assumption is load-bearing for the SmellNet-V dataset quality, the dense local alignment objective, and the derived saliency maps.

    Authors: We agree that the invariance assumption requires empirical support to fully validate the dataset construction. Although the assumption is motivated by the semantic consistency of odors (e.g., the odor of 'rose' remains similar regardless of the visual depiction), we acknowledge the lack of reported validation in the current manuscript. In the revised version, we will conduct and report a human study to assess pairing quality and consistency, including agreement scores, and discuss its implications for the method. revision: yes

  2. Referee: [§5 (Experiments)] The claimed 7% improvement in smell classification from smell alone is presented without error bars, statistical significance tests, dataset statistics, or ablations that isolate the contribution of the visual pairing versus the self-supervised objective. This makes it impossible to determine whether the lift is robust or an artifact of the synthetic construction.

    Authors: We concur that additional statistical rigor is necessary to substantiate the reported improvements. We will update §5 to include error bars from multiple experimental runs, statistical significance tests, detailed dataset statistics, and ablations that separate the effects of visual pairing from the self-supervised objective. This will clarify the robustness of the 7% gain. revision: yes

  3. Referee: [§4 (Method)] The dense local alignment loss and the procedure for extracting smell saliency maps from the learned representations are described only at a high level; no explicit equations or algorithmic details are provided for either component, preventing assessment of their technical novelty relative to prior cross-modal alignment methods.

    Authors: We appreciate the feedback on the need for greater technical specificity. In the revised manuscript, we will expand §4 to include explicit mathematical formulations for the dense local alignment loss and the saliency map extraction procedure, along with algorithmic pseudocode. This will facilitate direct comparison with existing cross-modal methods and highlight any novel aspects. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper's core steps consist of (1) constructing SmellNet-V via semantic-category pairing of existing smell samples with web images under an explicit invariance assumption, and (2) training a self-supervised model with dense local alignment on that dataset. No equations, fitted parameters, or predictions are described that reduce by construction to the inputs (no self-definitional loops, no fitted-input-called-prediction, no load-bearing self-citations, no uniqueness theorems, no ansatz smuggling). The reported 7% lift and cross-modal generalization are empirical outcomes of training rather than algebraic identities. The derivation chain is therefore self-contained against external benchmarks and receives the default non-circularity finding.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on one domain assumption about odor invariance; no free parameters or new invented entities are described.

axioms (1)
  • domain assumption Odor identity is largely invariant to visual transformations within a semantic category
    This premise is used to justify automatic pairing of smell samples with web images.

pith-pipeline@v0.9.1-grok · 5716 in / 1207 out tokens · 19956 ms · 2026-06-26T05:28:12.703472+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

54 extracted references · 5 linked inside Pith

  1. [1]

    arXiv preprint arXiv:1906.02569 (2019)

    Abid, A., Abdalla, A., Abid, A., Khan, D., Alfozan, A., Zou, J.: Gradio: Hassle- free sharing and testing of ml models in the wild. arXiv preprint arXiv:1906.02569 (2019)

    Pith/arXiv arXiv 1906

  2. [2]

    Scientific reports (2022)

    Achebouche,R.,Tromelin,A.,Audouze,K.,Taboureau,O.:Applicationofartificial intelligencetodecodetherelationshipsbetweensmell,olfactoryreceptorsandsmall molecules. Scientific reports (2022)

    2022

  3. [3]

    In: ICCV (2017)

    Arandjelovic, R., Zisserman, A.: Look, listen and learn. In: ICCV (2017)

    2017

  4. [4]

    In: NeurIPS (2016)

    Aytar, Y., Vondrick, C., Torralba, A.: Soundnet: Learning sound representations from unlabeled video. In: NeurIPS (2016)

    2016

  5. [5]

    arXiv preprint arXiv:1607.06450 (2016)

    Ba, J.L., Kiros, J.R., Hinton, G.E.: Layer normalization. arXiv preprint arXiv:1607.06450 (2016)

    Pith/arXiv arXiv 2016

  6. [6]

    In: ICLR (2026)

    Carion, N., Gustafson, L., Hu, Y.T., Debnath, S., Hu, R., Suris, D., Ryali, C., Alwala,K.V.,Khedr,H.,Huang,A.,etal.:Sam3:Segmentanythingwithconcepts. In: ICLR (2026)

    2026

  7. [7]

    Scientific Reports (2025)

    Castellotti, S., Soldo, M., Plank, T., Viva, M.M.D., Greenlee, M.W.: Visual search performance depends on the congruency of olfactory sensations. Scientific Reports (2025)

    2025

  8. [8]

    In: CVPR (2022)

    Cheng, B., Misra, I., Schwing, A.G., Kirillov, A., Girdhar, R.: Masked-attention mask transformer for universal image segmentation. In: CVPR (2022)

    2022

  9. [9]

    In: ICASSP (2023)

    Elizalde, B., Deshmukh, S., Ismail, M.A., Wang, H.: Clap: Learning audio concepts from natural language supervision. In: ICASSP (2023)

    2023

  10. [10]

    IJCV (2015)

    Everingham, M., Eslami, S.A., Van Gool, L., Williams, C.K., Winn, J., Zisserman, A.: The pascal visual object classes challenge: A retrospective. IJCV (2015)

    2015

  11. [11]

    Hugging Face Spaces (2025), https : / / huggingface

    Fancy Feast: Joycaption Watermark Detection. Hugging Face Spaces (2025), https : / / huggingface . co / spaces / fancyfeast / joycaption - watermark - detection, Accessed 24 June 2026

    2025

  12. [12]

    In: ICLR (2026)

    Feng, D., Dai, W., Li, C., Pernigo, A., Wen, Y., Liang, P.P.: Smellnet: A large-scale dataset for real-world smell recognition. In: ICLR (2026)

    2026

  13. [13]

    arXiv preprint arXiv:2512.08683 (2025)

    Fichtelmann, P., Westermayr, J.: Machine learning for smell: Ordinal odor strength prediction of molecular perfumery components. arXiv preprint arXiv:2512.08683 (2025)

    Pith/arXiv arXiv 2025

  14. [14]

    In: ICML (2024)

    Fu, L., Datta, G., Huang, H., Panitch, W.C.H., Drake, J., Ortiz, J., Mukadam, M., Lambeta, M., Calandra, R., Goldberg, K.: A touch, vision, and language dataset for multimodal alignment. In: ICML (2024)

    2024

  15. [15]

    In: CVPR (2023)

    Girdhar, R., El-Nouby, A., Liu, Z., Singh, M., Alwala, K.V., Joulin, A., Misra, I.: Imagebind: One embedding space to bind them all. In: CVPR (2023)

    2023

  16. [16]

    In: ICLR (2022) 16 S

    Gong, Y., Rouditchenko, A., Liu, A.H., Harwath, D., Karlinsky, L., Kuehne, H., Glass, J.: Contrastive audio-visual masked autoencoder. In: ICLR (2022) 16 S. Kim et al

    2022

  17. [17]

    chirp" fromthe

    Hamilton,M.,Zisserman,A.,Hershey,J.R.,Freeman,W.T.:Separatingthe"chirp" fromthe"chat":Self-supervisedvisualgroundingofsoundandlanguage.In:CVPR (2024)

    2024

  18. [18]

    In: ECCV (2018)

    Harwath, D., Recasens, A., Surís, D., Chuang, G., Torralba, A., Glass, J.: Jointly discovering visual objects and spoken words from raw sensory input. In: ECCV (2018)

    2018

  19. [19]

    Intel Corporation: CVAT: Computer Vision Annotation Tool (2025),https:// www.cvat.ai/, Accessed 24 June 2026

    2025

  20. [20]

    Current Research in Food Science (2025)

    Iwata, H.: Interpretable multitask deep learning models for odor perception based on molecular structure. Current Research in Food Science (2025)

    2025

  21. [21]

    In: ICML (2021)

    Jia, C., Yang, Y., Xia, Y., Chen, Y.T., Parekh, Z., Pham, H., Le, Q., Sung, Y.H., Li, Z., Duerig, T.: Scaling up visual and vision-language representation learning with noisy text supervision. In: ICML (2021)

    2021

  22. [22]

    arXiv preprint arXiv:2601.19561 (2026)

    Kang,D.,Kim,J.,Park,J.,Lee,K.,Choi,J.W.,So,J.:Aromma:Unifyingolfactory embeddings for single molecules and mixtures. arXiv preprint arXiv:2601.19561 (2026)

    arXiv 2026

  23. [23]

    The Home Magazine (1934),https://www.afb

    Keller, H.: A neglected treasure. The Home Magazine (1934),https://www.afb. org/HelenKellerArchive?a=d&d=A-HK02-B225-F02-024

    1934

  24. [24]

    In: CVPR (2026)

    Kim, S., Lee, S., Ryu, H., Chung, J.S., Senocak, A.: Seeing through touch: Tactile- driven visual localization of material regions. In: CVPR (2026)

    2026

  25. [25]

    In: ICCV (2023)

    Kirillov, A., Mintun, E., Ravi, N., Mao, H., Rolland, C., Gustafson, L., Xiao, T., Whitehead, S., Berg, A.C., Lo, W.Y., et al.: Segment anything. In: ICCV (2023)

    2023

  26. [26]

    Science (2023)

    Lee, B.K., Mayhew, E.J., Sanchez-Lengeling, B., Wei, J.N., Qian, W.W., Little, K.A., Andres, M., Nguyen, B.B., Moloy, T., Yasonik, J., et al.: A principal odor map unifies diverse tasks in olfactory perception. Science (2023)

    2023

  27. [27]

    In: ICML (2023)

    Li, J., Li, D., Savarese, S., Hoi, S.: Blip-2: Bootstrapping language-image pre- training with frozen image encoders and large language models. In: ICML (2023)

    2023

  28. [28]

    In: ICML (2022)

    Li, J., Li, D., Xiong, C., Hoi, S.: Blip: Bootstrapping language-image pre-training for unified vision-language understanding and generation. In: ICML (2022)

    2022

  29. [29]

    In: CVPR (2022)

    Lüddecke, T., Ecker, A.: Image segmentation using text and image prompts. In: CVPR (2022)

    2022

  30. [30]

    arXiv preprint arXiv:2405.16108 (2024)

    Lyu, Y., Zheng, X., Kim, D., Wang, L.: Omnibind: Teach to build unequal-scale modality interaction for omni-bind of all. arXiv preprint arXiv:2405.16108 (2024)

    arXiv 2024

  31. [31]

    Chemical senses (2006)

    Mainland, J., Sobel, N.: The sniff is part of the olfactory percept. Chemical senses (2006)

    2006

  32. [32]

    Expert systems with applications (2019)

    Mueller, P., Salminen, K., Nieminen, V., Kontunen, A., Karjalainen, M., Isokoski, P., Rantala, J., Savia, M., Väliaho, J., Kallio, P., et al.: Scent classification by k nearest neighbors using ion-mobility spectrometry measurements. Expert systems with applications (2019)

    2019

  33. [33]

    In: ECCV (2024)

    Naeem, M.F., Xian, Y., Zhai, X., Hoyer, L., Van Gool, L., Tombari, F.: Silc: Im- proving vision language pretraining with self-distillation. In: ECCV (2024)

    2024

  34. [34]

    In: ECCV (2018)

    Owens, A., Efros, A.A.: Audio-visual scene analysis with self-supervised multisen- sory features. In: ECCV (2018)

    2018

  35. [35]

    arXiv preprint arXiv:2511.20544 (2025)

    Ozguroglu, E., Liang, J., Liu, R., Chiquier, M., DeTienne, M., Qian, W.W., Horowitz, A., Owens, A., Vondrick, C.: New york smells: A large multimodal dataset for olfaction. arXiv preprint arXiv:2511.20544 (2025)

    arXiv 2025

  36. [36]

    In: ICML (2021) See & Sniff 17

    Radford, A., Kim, J.W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., et al.: Learning transferable visual models from natural language supervision. In: ICML (2021) See & Sniff 17

    2021

  37. [37]

    Neuroscience & Biobehavioral Reviews (2021)

    Raithel, C.U., Gottfried, J.A.: Using your nose to find your way: Ethological com- parisons between human and non-human species. Neuroscience & Biobehavioral Reviews (2021)

    2021

  38. [38]

    In: CVPR (2025)

    Ryu, H., Kim, S., Chung, J.S., Senocak, A.: Seeing speech and sound: Distinguish- ing and locating audio sources in visual scenes. In: CVPR (2025)

    2025

  39. [39]

    arXiv preprint arXiv:1910.10685 (2019)

    Sanchez-Lengeling, B., Wei, J.N., Lee, B.K., Gerkin, R.C., Aspuru-Guzik, A., Wiltschko, A.B.: Machine learning for scent: Learning generalizable perceptual representations of small molecules. arXiv preprint arXiv:1910.10685 (2019)

    arXiv 1910

  40. [40]

    In: CVPR (2018)

    Senocak, A., Oh, T.H., Kim, J., Yang, M.H., Kweon, I.S.: Learning to localize sound source in visual scenes. In: CVPR (2018)

    2018

  41. [41]

    In: ICCV (2023)

    Senocak, A., Ryu, H., Kim, J., Oh, T.H., Pfister, H., Chung, J.S.: Sound source localization is all about cross-modal alignment. In: ICCV (2023)

    2023

  42. [42]

    Senocak, A., Ryu, H., Kim, J., Oh, T.H., Pfister, H., Chung, J.S.: Toward in- teractive sound source localization: Better align sight and sound! IEEE TPAMI (2025)

    2025

  43. [43]

    arXiv preprint arXiv:2508.10104 (2025)

    Siméoni, O., Vo, H.V., Seitzer, M., Baldassarre, F., Oquab, M., Jose, C., Khali- dov, V., Szafraniec, M., Yi, S., Ramamonjisoa, M., et al.: Dinov3. arXiv preprint arXiv:2508.10104 (2025)

    Pith/arXiv arXiv 2025

  44. [44]

    arXiv preprint arXiv:2601.03267 (2025)

    Singh, A., Fry, A., Perelman, A., Tart, A., Ganesh, A., El-Kishky, A., McLaughlin, A., Low, A., Ostrow, A., Ananthram, A., et al.: Openai gpt-5 system card. arXiv preprint arXiv:2601.03267 (2025)

    Pith/arXiv arXiv 2025

  45. [45]

    Nature communications (2024)

    Sung,S.H.,Suh,J.M.,Hwang,Y.J.,Jang,H.W.,Park,J.G.,Jun,S.C.:Data-centric artificial olfactory system based on the eigengraph. Nature communications (2024)

    2024

  46. [46]

    In: ICML (2019)

    Tran, N., Kepple, D., Shuvaev, S., Koulakov, A.: Deepnose: Using artificial neural networks to represent the space of odorants. In: ICML (2019)

    2019

  47. [47]

    In: NeurIPS (2017)

    Vaswani,A.,Shazeer,N.,Parmar,N.,Uszkoreit,J.,Jones,L.,Gomez,A.N.,Kaiser, L., Polosukhin, I.: Attention is all you need. In: NeurIPS (2017)

    2017

  48. [48]

    Neuron (2011)

    Wachowiak, M.: All in a sniff: olfaction as a model for active sensing. Neuron (2011)

    2011

  49. [49]

    In: CVPR (2018)

    Wu, Z., Xiong, Y., Yu, S., Lin, D.: Unsupervised feature learning via non- parametric instance-level discrimination. In: CVPR (2018)

    2018

  50. [50]

    In: CVPR (2024)

    Yang, F., Feng, C., Chen, Z., Park, H., Wang, D., Dou, Y., Zeng, Z., Chen, X., Gangopadhyay, R., Owens, A., et al.: Binding touch to everything: Learning unified multimodal tactile representations. In: CVPR (2024)

    2024

  51. [51]

    In: NeurIPS - Datasets and Benchmarks Track (2022)

    Yang, F., Ma, C., Zhang, J., Zhu, J., Yuan, W., Owens, A.: Touch and go: Learning from human-collected vision and touch. In: NeurIPS - Datasets and Benchmarks Track (2022)

    2022

  52. [52]

    In: ICCV (2023)

    Yang, F., Zhang, J., Owens, A.: Generating visual scenes from touch. In: ICCV (2023)

    2023

  53. [53]

    seeds” (rather than “leaves

    Zhai, X., Mustafa, B., Kolesnikov, A., Beyer, L.: Sigmoid loss for language image pre-training. In: ICCV (2023) 18 S. Kim et al. – Supplementary Material – See & Sniff: Learning Visuo-Olfactory Representations The contents in this supplementary material are as follows: 7 Details onSmellNet-V.......................................... 18 8 Implementation De...

    2023

  54. [54]

    Photorealism: ‘a photo of real, high-quality {category}’ ‘a drawing of {category}’, ‘an illustration of {category}’, ‘a cartoon of {category}’3

    Category Consistency: ‘a photo of real, high-quality {category}’ ‘a photo of something else entirely’2. Photorealism: ‘a photo of real, high-quality {category}’ ‘a drawing of {category}’, ‘an illustration of {category}’, ‘a cartoon of {category}’3. Valid Object State: ‘a photo of fresh, good-quality {category}’ ‘a photo of rotten, spoiled, or moldy {categ...

    arXiv 2072