Pith. sign in

REVIEW 6 major objections 4 minor 38 references

Synthesizing Images on Perceptual Boundaries of ANNs for Uncovering Human Perceptual Variability on Facial Expressions

T0 review · 6 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Images that an ANN finds ambiguous between two expressions are the faces on which humans disagree, and fine-tuning on human choices makes the network predict individual perception.

desk verdict A genuinely useful generation pipeline and a substantial behavioral dataset, but the central claim of heightened human uncertainty is not yet demonstrated without baseline comparisons and out-of-sample evaluation. read the letter →

arxiv 2507.14549 v1 pith:KIT5H6WY submitted 2025-07-19 cs.CV cs.CY

Haotian Deng , Chi Zhang , Chen Wei , Quanying Liu This is my paper
classification cs.CVcs.CY
keywords facialexpressionrecognitionperceptualvariabilityANNdecisionboundarydiffusionmodelhuman-AIalignmentemotionperceptionvardataset
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

Facial expressions that sit on the decision boundary of an ANN classifier, where the network cannot commit to one emotion, are also the expressions on which human observers disagree most. To test this, the paper builds a perceptual boundary sampling method that generates faces via a two-stage diffusion process guided by an uncertainty loss, then filters the outputs through classifier activation thresholds. The resulting varEmotion dataset, with 1,678 images and 22,450 judgments from 66 participants, shows that these ANN-confusing faces provoke high entropy in human emotion choices, with the guiding emotion pair attracting roughly 80% of images into success or bias categories. Fine-tuning three classifiers on human behavioral data improves both group-level and individual-level prediction of human choices, and raises the correlation between model and human judgment entropy substantially. If correct, ANN decision boundaries are a practical source of emotionally ambiguous stimuli, and behavioral fine-tuning is a route to personalized emotion recognition.

What carries the argument

The load-bearing mechanism is uncertainty guidance during diffusion sampling. In a first stage, a diffusion model denoises image embeddings while being steered by the loss $loss(x,y) = -p(y|x)q(y)$, which raises the classifier's probability for two target emotions and lowers it for the rest; the sampling step is $x_{t-1} = \mathrm{DDPM}^{-}(x_t) - \gamma\nabla_{x_t} loss(x_t,y)$ with $\gamma = 0.5$. The resulting embeddings are rendered into images by a second-stage text-to-image model, and candidates are kept only when both target emotions' activations exceed their 75th percentiles on the RAF-DB dataset. Human choices on the surviving images form the varEmotion dataset. Then an MLP head on each of three network architectures is fine-tuned on mixed group and individual behavioral data, which is the step that transfers human variability back into the model.

What would settle it

Ask a separate group of raters to score naturalness of boundary-generated and non-boundary generated faces; if boundary images are rated less natural and naturalness predicts disagreement entropy, the central claim is an artifact. Alternatively, match boundary and non-boundary faces for naturalness and compare human disagreement entropy; no difference would falsify the claim that ANN boundaries map to human perceptual variability.

Watch

Extended reading notes

Core claim

The central discovery is an empirical correspondence: stimuli deliberately placed on ANN classification boundaries transfer their ambiguity to human observers. The paper claims that images whose ANN activations are uncertain between two emotions, such as anger versus fear, yield human choice distributions with high entropy; the guiding emotion pair still dominates the choices for roughly 80% of images, so the disagreement is structured rather than random. It further claims that fine-tuning a network on human trial data moves its predictions toward both the group distribution and each individual's own patterns, with individual-level fine-tuning adding about 1 to 3.5% accuracy over group-level fine-tuning and raising the Spearman correlation between model and human judgment entropy from 0.26 to 0.85 for one architecture. On the paper's terms, ANN decision boundaries and human perceptual boundaries are aligned closely enough that one can be used to find the other.

Load-bearing premise

The load-bearing assumption is that the generated boundary images look like real, natural faces, so the human disagreement they trigger comes from emotional ambiguity rather than from the images being artificial or uncanny, but the pipeline's only filter is an ANN activation threshold with no human naturalness check.

Editorial extensions

If this is right

  • Boundary sampling can produce on-demand stimuli for any emotion pair that confuse both the ANN and human observers, with nearly 80% of generated images falling into the paper's success or bias categories.
  • Fine-tuning on human behavioral data does not degrade standard benchmark accuracy on RAF-DB while improving prediction on the ambiguous varEmotion images.
  • Individual-level fine-tuning with a relatively small number of trials per person outperforms group-level fine-tuning on predicting that person's choices.
  • Architecture matters: the network with the largest group-level gain improves by 35% on varEmotion, while the smallest gain is 5%, showing that some models can fit human perceptual boundaries better than others.
  • Model uncertainty becomes human-like after fine-tuning: for one architecture the Spearman correlation between model and human choice entropy rises from 0.26 to 0.85, meaning the model reproduces how unsure people are, not just what they choose.

Reading between the lines

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

  • Going beyond the paper: if the boundary-to-variability link is causal rather than correlational, the same two-stage diffusion pipeline should generate ambiguous stimuli in any domain where a classifier's softmax boundary can be defined, not just facial expressions.
  • Going beyond the paper: because the filtering step uses only ANN activations, the decisive control is to compare boundary-generated faces against non-boundary generated faces matched for human-rated naturalness; if disagreement entropy no longer differs, the shared-boundary claim would reduce to an image-quality effect.
  • Going beyond the paper: since individual fine-tuning worked with only a few hundred trials per participant, the protocol could become a lightweight calibration step for personalized affective computing rather than requiring large per-person datasets.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

6 major / 4 minor

Summary. The paper proposes a pipeline for synthesizing facial expression images on the decision boundaries of ANN classifiers, using a two-stage diffusion process with an 'uncertainty guidance' loss and a 75th-percentile activation filter. The resulting 1,678 images were shown to 66 human participants in 22,450 trials, forming the varEmotion dataset. The authors report that these images elicit high perceptual variability in humans (entropy distributions and an approximately 80% success-plus-bias rate), and that fine-tuning CLIP, DAN, and ResEmoNet on group-level and individual-level behavioral data improves their accuracy on varEmotion and varEmotion-i and increases the Spearman correlation between model and human entropy. The paper concludes that ANN decision boundaries are a systematic source of ambiguous stimuli for humans and that behavioral fine-tuning can align models with individual-level perceptual patterns.

Significance. If the central claim were fully supported, the paper would make a useful contribution: it would connect ANN decision boundaries to human perceptual variability, provide a new behavioral dataset (varEmotion), and show that human choice data can personalize emotion classifiers. The dataset collection effort is substantial, and the cross-architecture comparison (CLIP, DAN, ResEmoNet) gives the fine-tuning results some breadth. However, the current evidence is suggestive rather than conclusive: the 'heightened uncertainty' claim lacks a baseline, the fine-tuning evaluations appear to be on training data, and the generation/filtering loop uses the same ANN whose shared boundary with humans is being asserted. These issues are fixable within the manuscript's scope, but they are load-bearing for the abstract and conclusion.

major comments (6)
  1. [Sec. IV-A, Fig. 7(a)] The claim that ANN-boundary images provoke 'heightened' perceptual uncertainty is not supported without a baseline: the entropy distribution is reported only for generated boundary images, and the roughly 80% success-plus-bias rate in Fig. 4(b) counts the 'bias' outcome (all subjects choosing one target) as a positive result even though that outcome reflects agreement rather than variability. I recommend comparing the entropy of these images with natural faces, off-boundary generated faces, or morph continua, and reporting success and bias rates separately.
  2. [Sec. V-A, Fig. 5(a)] The fine-tuning evaluation does not separate training from test images: GroupNet and IndivNet are fine-tuned on varEmotion and varEmotion-i and then evaluated on those same datasets, so the accuracy improvements are expected from fitting the training labels. Please report performance on held-out images using image-level or subject-level cross-validation, and include BaseNet on the same held-out splits.
  3. [Sec. V-B, Fig. 5(c)] The entropy correlation improvement from ρ = 0.26 to ρ = 0.85 is computed on varEmotion images that were used for group fine-tuning; a model trained to reproduce human choices on those images would be expected to have correlated entropy. Please compute the Spearman correlation on held-out images and report confidence intervals or p-values.
  4. [Sec. III-C, Eq. (3)] The filtering criterion uses the same ANN whose decision boundaries are being probed, so the selected images are ambiguous for that specific classifier by construction; this does not by itself establish that the ambiguity is shared with humans. A control with a different classifier architecture, or a comparison of filtered versus unfiltered generated images, is needed to rule out selection artifacts.
  5. [Sec. III-B and Introduction] The manuscript asserts that the generated images retain 'photorealistic authenticity' but provides no human naturalness rating or artifact check. If the boundary images are uncanny or low-quality, the elevated human disagreement could reflect image unnaturalness rather than shared emotional ambiguity. A brief naturalness rating experiment or a comparison with real or morphed faces would resolve this.
  6. [Sec. III-B, Eq. (1)] The uncertainty loss is underspecified: q(y) is never defined, and the notation switches between p(y|x) and p(y). If q(y) is a distribution over the two target emotions, the product form -p(y|x) q(y) is not a standard objective and its optimization behavior is unclear; please define the exact target distribution (e.g., q = 0.5 for each target emotion) and report sensitivity to the guidance strength γ.
minor comments (4)
  1. [Fig. 7 caption] The caption refers to a 'Digit recognition task,' but the experiment is a facial expression recognition task; please correct this.
  2. [Sec. IV] Entropy is estimated from roughly 13 judgments per image on average across six categories, which can produce biased entropy estimates; please report bias-corrected entropy or bootstrap confidence intervals.
  3. [Sec. III-A and III-B] The section headings appear mismatched: Sec. III-A is titled 'Generating Images on ANN perceptual boundary' but contains no method details, while Sec. III-B is titled 'Facial expression recognition experiment' but describes the generation procedure; please reorganize the subsection structure.
  4. [References] Reference [8] is cited as evidence of 'remarkable accuracy' in facial expression recognition, but the listed work is about family interaction and appears unrelated; please verify and replace this citation.

Circularity Check

2 steps flagged · score 6.0 of 10

The human-ANN uncertainty link is tested with independent behavioral data, but the model-alignment 'predictions' are reported on the very varEmotion data used for fine-tuning, making the alignment evidence fit quality rather than prediction.

  1. fitted input called prediction [Section V-A (Model Finetuning For Human Alignment) and Section V-B (Alignment Analysis), Fig. 5]
    "For group-level fine-tuning, the original classifier models were trained on mixed RAF-DB, varEmotion datasets."

    The group-level alignment claim is then supported by accuracy on varEmotion: 'On the varEmotion dataset, both GroupNet and IndivNet show improved prediction accuracy compared to BaseNet' and by the entropy correlation increase for DAN from rho=0.26 to rho=0.85. Because the classifier weights were fine-tuned on the varEmotion human labels, these numbers measure how well the model fits its own training labels, not how well a boundary-based model predicts held-out human variability. The 'prediction' of human perceptual patterns is therefore the fitted input itself.

  2. fitted input called prediction [Section V-A (Model Finetuning For Human Alignment) and Section V-B (Alignment Analysis), Fig. 5]
    "Moreover, on the individual dataset varEmotion-i, IndivNet, which is fine-tuned using individual data, demonstrates an average prediction accuracy improvement of 3% over the group model GroupNet."

    IndivNet is the group model further fine-tuned on varEmotion-i, the same individual-level behavioral data on which the 3% improvement is reported. The reported gain is a measure of how well the individual fine-tuning procedure memorized or fit that individual's choices; it is not an out-of-sample prediction of individual perceptual variability. This is the individual-level analogue of the group-level fitted-input issue.

full rationale

The central empirical hypothesis—that ANN-ambiguous facial stimuli evoke divergent human emotion judgments—is not circular by construction: varEmotion human responses were collected independently of the ANN's labels, and the human entropy distribution, reaction times, and success/bias rates could have failed to show disagreement. The generation pipeline's use of the same RAF-DB-trained MLP for guidance and filtering (Eqs. 1-3) is a selection criterion, not a definitional identification of human uncertainty; a proper baseline would be needed to support 'heightened' uncertainty, but that is a control/artifact concern rather than circularity. The paper's self-citations to CoCoG [1], [2]/[36] are methodological (first-stage embedding diffusion and diffusion regularizers) and are not used to prove the main claim. The substantial circularity is confined to Section V: both group-level and individual-level 'alignment' results are evaluated on the varEmotion / varEmotion-i data used to fine-tune the models, so the accuracy gains and entropy correlations (e.g., DAN rho 0.26 -> 0.85) are fit-quality statistics. Since the paper's novel dataset contribution and the human-ANN uncertainty observation remain independently grounded, the overall circularity is partial rather than total.

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

The central machinery adds one hand-chosen guidance strength, one post hoc filter threshold per emotion, and hand-defined evaluation thresholds. The key domain assumption is that ANN boundary directions correspond to human ambiguity, and that generated images are natural enough not to confound responses.

free parameters (4)
  • guidance strength gamma = 0.5
    Hand-chosen in Eq. 2 for first-stage diffusion; no ablation, affects how strongly samples are pushed to the ANN boundary.
  • k_emotion activation thresholds = 75th percentile of RAF-DB activation per emotion
    Used in Eq. 3 to filter generated images; the percentile choice is post hoc and not validated against human ambiguity.
  • success/bias thresholds = min(p1,p2)>0.25 and p1+p2>0.6
    Hand-chosen thresholds define the 80% success+bias headline in Sec. IV; no justification is given.
  • fine-tuning data ratios and epochs = 2:1 mixture, 4:1 split, lr=1e-4, 15 epochs
    Training choices in Sec. V-A affect reported gains; not varied systematically.
assumptions (5)
  • domain assumption A RAF-DB-trained MLP's decision boundaries in CoCoG embedding space are meaningful proxies for human perceptual ambiguity.
    This is the paper's core hypothesis, imported from Refs [1],[10],[11]; the paper provides no independent validation that the boundary directions match human perceptual dimensions.
  • domain assumption The two-stage diffusion generation plus activation filtering yields photorealistic faces, so participant choices reflect emotion ambiguity rather than image artifacts.
    Stated in Sec. III-B/III-C with 'generative uncertainty constraints', but no human naturalness ratings are collected.
  • domain assumption The six basic emotion categories (surprise, fear, disgust, happiness, sadness, anger) are the appropriate exhaustive label set.
    Adopted from Ekman-style literature [29]-[33]; affects all entropy and accuracy measures.
  • standard math Reverse diffusion steps DDP M-(x_t) are valid stochastic updates with standard classifier guidance.
    Uses the U-ViT sampler from [38]; treated as background.
  • domain assumption Cross-entropy fine-tuning on human choice labels is a valid objective for aligning model predictions with human perception.
    Optimization design in Sec. V-A; treated as standard.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Synthesizing Images on Perceptual Boundaries of ANNs for Uncovering Human Perceptual Variability on Facial Expressions." pith.science (2026). https://pith.science/paper/KIT5H6WY

@misc{pith2026250714549,
  author       = {Pith},
  title        = {Pith review of: Synthesizing Images on Perceptual Boundaries of ANNs for Uncovering Human Perceptual Variability on Facial Expressions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KIT5H6WY}},
  note         = {Machine review of arXiv:2507.14549}
}
read the original abstract

A fundamental challenge in affective cognitive science is to develop models that accurately capture the relationship between external emotional stimuli and human internal experiences. While ANNs have demonstrated remarkable accuracy in facial expression recognition, their ability to model inter-individual differences in human perception remains underexplored. This study investigates the phenomenon of high perceptual variability-where individuals exhibit significant differences in emotion categorization even when viewing the same stimulus. Inspired by the similarity between ANNs and human perception, we hypothesize that facial expression samples that are ambiguous for ANN classifiers also elicit divergent perceptual judgments among human observers. To examine this hypothesis, we introduce a novel perceptual boundary sampling method to generate facial expression stimuli that lie along ANN decision boundaries. These ambiguous samples form the basis of the varEmotion dataset, constructed through large-scale human behavioral experiments. Our analysis reveals that these ANN-confusing stimuli also provoke heightened perceptual uncertainty in human participants, highlighting shared computational principles in emotion perception. Finally, by fine-tuning ANN representations using behavioral data, we achieve alignment between ANN predictions and both group-level and individual-level human perceptual patterns. Our findings establish a systematic link between ANN decision boundaries and human perceptual variability, offering new insights into personalized modeling of emotional interpretation.

Figures

Figures reproduced from arXiv: 2507.14549 by the authors.

Figure 1
Figure 1. Overview of our paradigm. (a) Motivation: An example of perceptual variability. (b) Our approach consists of two main components: 1. Generating & labeling: Sampling images from ANN decision boundaries and using them in human behavioral experiments to construct the high-variability dataset varEmotion; 2. Predicting: Finetuning models with human behavioral data to align them with human perceptual variability at the gr… view at source ↗
Figure 2
Figure 2. Generating images to elicit human perceptual variability. (a) This example demonstrates how to generate embeddings by sampling from the perceptual boundaries of the expressions ‘anger’ and ‘fear’ in an ANN using the uncertainty guidance method. The goal of uncertainty guidance is to focus the ANN’s prediction of the embeddings on ‘anger’ and ‘fear’. The diffusion model follows the first stage generation process from… view at source ↗
Figure 3
Figure 3. Human facial expression recognition experiment procedure. In each round of the experiment, the participant will first see a cross at the center of the screen for 300 ms. Following this, a facial image containing a specific expression will be presented for 200 ms. Next, a choice page with six buttons will appear, and the participant is required to judge the expression of the face in the image just shown and select th… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Quantitative Analysis of varEmotion. (a) Examples of three guidance outcome:success, bias, failure. (b) Guidance outcome across the varEmotion dataset. The sum of success and bias rates approaches 80% . ANN variability can arouse human variability. To ex￾amine whether …
Figure 5
Figure 5. Figure 5: Human alignment results. (a) Accuracy of BaseNet, GroupNet and IndivNet on RAF-DB, varEmotion and varEmotion-i. On varEmotion, GroupCLIP and IndivCLIP improve 5% over baseNet, GroupDAN and IndivDAN improve 14%, GroupResEmoNet and IndivResEmoNet improve 35%. On varEmoti…
Figure 7
Figure 7. Figure 7: Behavioral results of the Digit recognition task. (a) The entropy distribution of human judgments on images primarily concentrates between 0.5 and 2.0, approximately following a normal distribution. (b) The time humans take to judge the images is concentrated between 5…
Figure 6
Figure 6. Figure 6: Sampling on perceptual boundaries. The perceptual space of ANN can be divided into four regions based on two classification axes. Taking the emotion pairs (fear, surprise) as an example. our objective is to generate images that induce uncertainty of ANN, as illustrated…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

38 extracted references · 30 canonical work pages

  1. [1]

    CoCoG: Controllable Visual Stimuli Generation based on Human Concept Representations

    C. Wei, J. Zou, D. Heinke, and Q. Liu, “Cocog: Controllable visual stimuli generation based on human concept representations,” arXiv preprint arXiv:2404.16482, 2024

    work page Pith review arXiv 2024

  2. [2]

    CoCoG-2: Controllable generation of visual stimuli for understanding human concept representation

    ——, “Cocog-2: Controllable generation of visual stimuli for understanding human concept representation,” arXiv preprint arXiv:2407.14949, 2024

    work page Pith review arXiv 2024

  3. [3]

    Human alignment of neural network representations,

    L. Muttenthaler, J. Dippel, L. Linhardt, R. A. Vandermeulen, and S. Kornblith, “Human alignment of neural network representations,” arXiv preprint arXiv:2211.01201 , 2022

    arXiv 2022

  4. [4]

    Dimensions underlying the representational alignment of deep neural networks with humans

    F. P. Mahner, L. Muttenthaler, U. Güçlü, and M. N. Hebart, “Dimensions underlying the representational alignment of deep neural networks with humans,” arXiv preprint arXiv:2406.19087 , 2024

    work page Pith review arXiv 2024

  5. [5]

    Revealing interpretable object representations from human behavior

    C. Y . Zheng, F. Pereira, C. I. Baker, and M. N. Hebart, “Revealing in- terpretable object representations from human behavior,” arXiv preprint arXiv:1901.02915, 2019

    work page Pith review arXiv 1901

  6. [6]

    Revealing the multidimensional mental representations of natural objects underlying human similarity judgements,

    M. N. Hebart, C. Y . Zheng, F. Pereira, and C. I. Baker, “Revealing the multidimensional mental representations of natural objects underlying human similarity judgements,” Nature human behaviour, vol. 4, no. 11, pp. 1173–1185, 2020

    work page 2020

  7. [7]

    Vice: Variational interpretable concept embed- dings,

    L. Muttenthaler, C. Y . Zheng, P. McClure, R. A. Vandermeulen, M. N. Hebart, and F. Pereira, “Vice: Variational interpretable concept embed- dings,” Advances in Neural Information Processing Systems, vol. 35, pp. 33 661–33 675, 2022

    work page 2022

  8. [8]

    The face of noncompliance in family interaction,

    M. H. Goodwin and H. Loyd, “The face of noncompliance in family interaction,” Text & Talk, vol. 40, no. 5, pp. 573–598, 2020

    work page 2020

Show all 38 references
  1. [9]

    Hierarchical text-conditional image generation with clip latents,

    A. Ramesh, P. Dhariwal, A. Nichol, C. Chu, and M. Chen, “Hierarchical text-conditional image generation with clip latents,” arXiv preprint arXiv:2204.06125, vol. 1, no. 2, p. 3, 2022

    2022 arXiv

  2. [10]

    Subtle adversarial image manipulations influence both human and machine perception,

    V . Veerabadran, J. Goldman, S. Shankar, B. Cheung, N. Papernot, A. Kurakin, I. Goodfellow, J. Shlens, J. Sohl-Dickstein, M. C. Mozer et al. , “Subtle adversarial image manipulations influence both human and machine perception,” Nature Communications , vol. 14, no. 1, p. 4933, 2023

    2023

  3. [11]

    Strong and precise modulation of human percepts via robustified anns,

    G. Gaziv, M. Lee, and J. J. DiCarlo, “Strong and precise modulation of human percepts via robustified anns,” Advances in Neural Information Processing Systems, vol. 36, 2024

    2024

  4. [12]

    Model metamers reveal divergent invariances between biological and artificial neural networks,

    J. Feather, G. Leclerc, A. M ˛ adry, and J. H. McDermott, “Model metamers reveal divergent invariances between biological and artificial neural networks,” Nature Neuroscience, vol. 26, no. 11, pp. 2017–2034, 2023

    2017

  5. [13]

    Controversial stimuli: Pitting neural networks against each other as models of human cognition,

    T. Golan, P. C. Raju, and N. Kriegeskorte, “Controversial stimuli: Pitting neural networks against each other as models of human cognition,” Proceedings of the National Academy of Sciences , vol. 117, no. 47, pp. 29 330–29 337, 2020

    2020

  6. [14]

    Testing the limits of natural language models for predicting human language judgements,

    T. Golan, M. Siegelman, N. Kriegeskorte, and C. Baldassano, “Testing the limits of natural language models for predicting human language judgements,” Nature Machine Intelligence , vol. 5, no. 9, pp. 952–964, 2023

    2023

  7. [15]

    Metamers of neural networks reveal divergence from human perceptual systems,

    J. Feather, A. Durango, R. Gonzalez, and J. McDermott, “Metamers of neural networks reveal divergence from human perceptual systems,” Advances in Neural Information Processing Systems , vol. 32, 2019

    2019

  8. [16]

    Measuring representational robustness of neural networks through shared invariances,

    V . Nanda, T. Speicher, C. Kolling, J. P. Dickerson, K. Gummadi, and A. Weller, “Measuring representational robustness of neural networks through shared invariances,” in International Conference on Machine Learning. PMLR, 2022, pp. 16 368–16 382

    2022

  9. [17]

    Do invariances in deep neural networks align with human perception?

    V . Nanda, A. Majumdar, C. Kolling, J. P. Dickerson, K. P. Gummadi, B. C. Love, and A. Weller, “Do invariances in deep neural networks align with human perception?” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 37, no. 8, 2023, pp. 9277–9285

    2023

  10. [18]

    Dreamsim: Learning new dimensions of human visual similarity using synthetic data,

    S. Fu, N. Tamir, S. Sundaram, L. Chai, R. Zhang, T. Dekel, and P. Isola, “Dreamsim: Learning new dimensions of human visual similarity using synthetic data,” arXiv preprint arXiv:2306.09344 , 2023

    2023 arXiv

  11. [19]

    Aligning machine and human visual representations across abstraction levels,

    L. Muttenthaler, K. Greff, F. Born, B. Spitzer, S. Kornblith, M. C. Mozer, K.-R. Müller, T. Unterthiner, and A. K. Lampinen, “Aligning machine and human visual representations across abstraction levels,” arXiv preprint arXiv:2409.06509 , 2024

    2024 arXiv

  12. [20]

    When does perceptual alignment benefit vision representations?

    S. Sundaram, S. Fu, L. Muttenthaler, N. Y . Tamir, L. Chai, S. Kornblith, T. Darrell, and P. Isola, “When does perceptual alignment benefit vision representations?” arXiv preprint arXiv:2410.10817 , 2024

    2024 arXiv

  13. [21]

    Adversarial counterfactual visual explanations,

    G. Jeanneret, L. Simon, and F. Jurie, “Adversarial counterfactual visual explanations,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 16 425–16 435

    2023

  14. [22]

    Advdiffuser: Natural adversarial example synthesis with diffusion models,

    X. Chen, X. Gao, J. Zhao, K. Ye, and C.-Z. Xu, “Advdiffuser: Natural adversarial example synthesis with diffusion models,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 4562–4572

    2023

  15. [23]

    Diffusion models for counterfac- tual explanations,

    G. Jeanneret, L. Simon, and F. Jurie, “Diffusion models for counterfac- tual explanations,” in Proceedings of the Asian Conference on Computer Vision, 2022, pp. 858–876

    2022

  16. [24]

    Diffusion- based visual counterfactual explanations–towards systematic quantitative evaluation,

    P. Vaeth, A. M. Fruehwald, B. Paassen, and M. Gregorova, “Diffusion- based visual counterfactual explanations–towards systematic quantitative evaluation,” arXiv preprint arXiv:2308.06100 , 2023

    2023 arXiv

  17. [25]

    Dreamr: Diffusion-driven counterfactual explanation for functional mri,

    H. Atakan Bedel and T. Çukur, “Dreamr: Diffusion-driven counterfactual explanation for functional mri,” arXiv e-prints, pp. arXiv–2307, 2023

    2023

  18. [26]

    Freedom: Training- free energy-guided conditional diffusion model,

    J. Yu, Y . Wang, C. Zhao, B. Ghanem, and J. Zhang, “Freedom: Training- free energy-guided conditional diffusion model,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 23 174–23 184

    2023

  19. [27]

    Elucidating the design space of classifier-guided diffusion generation,

    J. Ma, T. Hu, W. Wang, and J. Sun, “Elucidating the design space of classifier-guided diffusion generation,”arXiv preprint arXiv:2310.11311, 2023

    2023 arXiv

  20. [28]

    Guidance with spherical gaussian constraint for conditional diffusion,

    L. Yang, S. Ding, Y . Cai, J. Yu, J. Wang, and Y . Shi, “Guidance with spherical gaussian constraint for conditional diffusion,” arXiv preprint arXiv:2402.03201, 2024

    2024 arXiv

  21. [29]

    Facial action coding system,

    P. Ekman and W. V . Friesen, “Facial action coding system,” Environ- mental Psychology & Nonverbal Behavior , 1978

    1978

  22. [30]

    Universals and cultural variations in 22 emotional expressions across five cultures

    D. T. Cordaro, R. Sun, D. Keltner, S. Kamble, N. Huddar, and G. Mc- Neil, “Universals and cultural variations in 22 emotional expressions across five cultures.” Emotion, vol. 18, no. 1, p. 75, 2018

    2018

  23. [31]

    Emotional expression: Advances in basic emotion theory,

    D. Keltner, D. Sauter, J. Tracy, and A. Cowen, “Emotional expression: Advances in basic emotion theory,” Journal of nonverbal behavior , vol. 43, pp. 133–160, 2019

    2019

  24. [32]

    Facial expressions of emotion,

    D. Matsumoto, D. Keltner, M. N. Shiota, M. O’Sullivan, and M. Frank, “Facial expressions of emotion,” Handbook of emotions, vol. 3, pp. 211– 234, 2008

    2008

  25. [33]

    Dynamic facial expressions of emotion transmit an evolving hierarchy of signals over time,

    R. E. Jack, O. G. Garrod, and P. G. Schyns, “Dynamic facial expressions of emotion transmit an evolving hierarchy of signals over time,” Current biology, vol. 24, no. 2, pp. 187–192, 2014

    2014

  26. [34]

    Testing, explaining, and exploring models of facial expressions of emotions,

    L. Snoek, R. E. Jack, P. G. Schyns, O. G. Garrod, M. Mittenbühler, C. Chen, S. Oosterwijk, and H. S. Scholte, “Testing, explaining, and exploring models of facial expressions of emotions,” Science advances, vol. 9, no. 6, p. eabq8421, 2023

    2023

  27. [35]

    Generative adversarial networks,

    I. J. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y . Bengio, “Generative adversarial networks,” 2014. [Online]. Available: https://arxiv.org/abs/1406.2661

    2014 arXiv

  28. [36]

    Cocog-2: Controllable generation of visual stimuli for understanding human concept representation,

    C. Wei, J. Zou, D. Heinke, and Q. Liu, “Cocog-2: Controllable generation of visual stimuli for understanding human concept representation,” 2024. [Online]. Available: https://arxiv.org/abs/2407. 14949

    2024

  29. [37]

    Diffusion- based visual counterfactual explanations-towards systematic quantitative evaluation,

    P. Väth, A. M. Frühwald, B. Paassen, and M. Gregorova, “Diffusion- based visual counterfactual explanations-towards systematic quantitative evaluation,” in Joint European Conference on Machine Learning and Knowledge Discovery in Databases . Springer, 2023, pp. 120–135

    2023

  30. [38]

    All are worth words: A vit backbone for diffusion models,

    F. Bao, S. Nie, K. Xue, Y . Cao, C. Li, H. Su, and J. Zhu, “All are worth words: A vit backbone for diffusion models,” 2023. [Online]. Available: https://arxiv.org/abs/2209.12152

    2023 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.