Pith. sign in

REVIEW 2 major objections 5 minor 50 references

Advancing Generalization Across a Variety of Abstract Visual Reasoning Tasks

T0 review · 2 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A single architecture, the Pathways of Normalized Group Convolution model (PoNG), claims to generalize across Raven's Progressive Matrices and visual analogy benchmarks, synthetic and real, and reports top results in several…

desk verdict Solid AVR architecture paper whose headline gains are partly explained by ground-truth rule supervision; a matched-supervision comparison is needed before claiming the architecture generalizes. read the letter →

arxiv 2505.13391 v1 pith:P4ZY5AM6 submitted 2025-05-19 cs.AI cs.CVcs.LG

Miko{l}aj Ma{l}kiński , Jacek Mańdziuk This is my paper · ORCID
classification cs.AIcs.CVcs.LG
keywords abstractvisualreasoningRaven'sProgressiveMatricesout-of-distributiongeneralizationgroupconvolutionanalogytemporalcontextnormalizationauxiliaryrulesupervision
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

The paper introduces PoNG, a neural network for abstract visual reasoning puzzles such as Raven's Progressive Matrices, where the solver must pick the panel that completes a pattern. PoNG's claim is that its design, built around normalized group convolutions that process rows and row-pairs of the puzzle grid, transfers across many such tasks and generalizes to test puzzles whose rules were not seen in training. The reported numbers support the claim on several benchmarks: 80.3% on A/Color versus 70.0% for the best baseline, and 91.8% on real-world visual analogies versus 70.5% for Supervised Concat. A fair reader should take away that a single parameter-efficient architecture can top most baselines on both synthetic and real-world analogy reasoning, with the caveat that the training loss uses the puzzle's ground-truth rules as auxiliary supervision.

What carries the argument

The load-bearing component is the Pathways block inside each reasoning block, with four parallel pathways P1 through P4. P3 and P4 use the named group convolution and group-pair convolution operators: they split the stacked panel embeddings into groups corresponding to matrix rows or row pairs, apply a shared-weight 1D convolution to each group, then temporal context normalization (TCN), and sum the groups. TCN, taken from the paper's reference Webb et al. (2020), preserves relations between representations within a group while discarding absolute magnitude, which the ablation study ties to extrapolating rules to held-out attributes. The two rule heads, one aggregating all answers and one weighting answers by the predicted answer distribution, inject the known rule structure into training through binary cross-entropy terms.

What would settle it

Train RelBase or SCL under the identical loss, including the two rule heads with $\beta=25$ and $\gamma=5$, on A-I-RAVEN; if A/Color accuracy rises from about 67% to about 80%, then the auxiliary rule supervision, not the group-convolution pathways, explains the gain. If the gap persists, the pathways carry the generalization.

Watch

Extended reading notes

Core claim

PoNG is a two-stage model: a panel encoder turns each image into a vector, and a reasoner scores each candidate answer. The reasoner stacks panel embeddings with a learned position embedding and passes them through three blocks, each containing four parallel pathways: pointwise mixing (P1), local 1D convolutions (P2), group convolution (P3), and group-pair convolution (P4), with the last two augmented by temporal context normalization (TCN). Grouping is chosen so that P3 operates on rows of the matrix and P4 on pairs of rows. Three prediction heads produce the answer score and two rule predictions; the loss combines cross-entropy for the answer with binary cross-entropy for the two rule heads, with coefficients $\beta=25$ and $\gamma=5$. The central claim is that this combination generalizes across benchmark families: on A/Color, PoNG scores 80.3% against a 70.0% best baseline; on A/Color-D3 it leads by 15.6 percentage points; on VASR difficult distractors it reaches 70.5% best-of-three versus 54.9% for Supervised Concat; and on PGM it averages 57.3%, second to DRNet's 58.3%. The paper presents this as evidence of strong generalization from a single architecture rather than task-specific design.

Load-bearing premise

The reported advantage over baselines assumes that the only meaningful difference is the architecture, not the extra ground-truth rule labels PoNG uses in its auxiliary loss.

Editorial extensions

If this is right

  • On Raven-style benchmarks, PoNG's strongest gains appear exactly where the test distribution swaps in rules unseen at training, so the row- and row-pair grouping is doing generalization work rather than memorization.
  • On real-world VASR, the frozen perception backbone is the same across models, so the 70.5% versus 54.9% difficult-distractor gap isolates the reasoner's contribution on top of shared perception.
  • Because the rule-head losses are removable and removing both drops I-RAVEN accuracy from 95.9% to 79.7%, below RelBase and SCL, deploying PoNG on new benchmarks should preserve the auxiliary rule labels or expect the generalization gain to shrink.
  • The parameter count of 3.1M is far below several baselines (for example DRNet at 24.7M and SRAN at 45.7M), so the reported gains are not bought by scale.
  • The paper states that the pathways block is a generic module over sets of vector representations, implying the same reasoner could be lifted from image puzzles to other set-structured inputs without architectural change.

Reading between the lines

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

  • The paper does not report whether the baseline models were trained with ground-truth rule supervision. If they were not, a fair test of PoNG's architectural contribution is to give the same rule-supervision loss to the baselines; the ablation where both rule heads are removed drops I-RAVEN accuracy from 95.9% to 79.7%, suggesting part of the reported gap could be supervision rather than convolutio
  • Because the ablation without both rule heads falls below RelBase and SCL, the comparison that isolates the architecture alone would be PoNG without rule heads against those baselines; on I-RAVEN that comparison would not favor PoNG.
  • The row- and row-pair grouping is a strong inductive bias for grid puzzles. A testable extension is to run PoNG on a non-grid set-reasoning task, such as text or graph analogies, where grouping into rows is meaningless; if performance degrades sharply, the claimed generality of the pathways block would need qualification.
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

2 major / 5 minor

Summary. The paper proposes PoNG, a neural architecture for abstract visual reasoning that combines group convolution, normalization, and parallel pathways. The model is evaluated on four RPM-style benchmarks (I-RAVEN, I-RAVEN-Mesh, A-I-RAVEN, PGM) and two visual analogy datasets (VAP, VASR), reporting accuracy improvements over 13 baselines in several out-of-distribution regimes. The training loss includes two auxiliary heads that predict ground-truth rule vectors r, with weights β=25 and γ=5. An ablation removing both heads shows substantial performance drops. The paper provides public code and fixed random seeds.

Significance. If the reported gains were attributable to the architecture alone, PoNG would be a strong contribution to visual reasoning, especially for out-of-distribution generalization across synthetic and real-world tasks. Strengths include reproducible code, transparent ablations, and evaluation on multiple benchmarks. However, the significance is substantially weakened by the unequal supervision between PoNG and baselines: the rule-label auxiliary loss is not accounted for in the comparisons, and the paper's own ablation shows that without it PoNG underperforms several baselines. The claims need to be re-framed or re-validated with a controlled comparison.

major comments (2)
  1. [Section 3, Eq. (1); Tables 1 and 6] The joint loss includes β·BCE(ζ(b_{r1}), r) + γ·BCE(ζ(b_{r2}), r) with β=25, γ=5, where r is the ground-truth multi-hot rule vector. The paper never states whether the 13 baselines received rule-label supervision, and the standard implementations of WReN, RelBase, SCL, DRNet, and PredRNet are trained only on answer labels. This confounds the architecture's contribution with access to exact abstract rules. Table 6 shows the magnitude: removing both rule heads (β=γ=0) lowers I-RAVEN accuracy from 95.9% to 79.7% and I-RAVEN-Mesh from 89.3% to 32.7%. The authors should disclose this supervision difference and provide a controlled comparison, e.g., by training baselines with the same rule-head loss or by comparing PoNG without rule heads on the same terms.
  2. [Section 4.3, Tables 1 and 6] The answer-only variant (β=γ=0) is the appropriate architecture-only baseline. Under that ablation, PoNG achieves 79.7% on I-RAVEN and 32.7% on I-RAVEN-Mesh, which are below RelBase (89.6%, 84.9%) and SCL (83.4%, 80.9%). Consequently, the claim in the abstract and Section 4.3 that PoNG outperforms the existing literature methods is not supported for the architecture alone. The central claim needs to be qualified to the full model with auxiliary rule supervision, or the experiments need to establish that the architecture is competitive without that supervision.
minor comments (5)
  1. [Section 3, last paragraph] The dimensionality dr is specified for I-RAVEN, A-I-RAVEN, I-RAVEN-Mesh, PGM, and VAP, but not for VASR; please clarify whether the rule heads are used for VASR and, if so, how the rule vector r is defined for real-world images.
  2. [Section 4.3, Table 5] PoNG results on VASR are reported as best-of-3 and compared to baselines from the original paper, which may be single runs; please report mean±std as in other tables and state whether the baselines used multiple seeds.
  3. [Section 4.2 and Table 1] The column I-RAVEN† reports results by the original model authors with model-specific setups, whereas other columns use the authors' own experimental setup; please clarify whether the comparison is apples-to-apples in terms of data splits and hyperparameter tuning.
  4. [Section 4.1] I-RAVEN-Mesh was originally introduced for transfer learning; using it for standard supervised learning is a protocol choice and should be explicitly justified or noted in the experimental setup.
  5. [Appendix B, Table 11] The error analysis is informative, but the table reports aggregated calibration metrics without confidence intervals; providing standard deviations or significance tests would strengthen the claim that TCN helps rule generalization.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: PoNG's results are empirical held-out accuracies with a disclosed loss and ablations; self-authored benchmarks and auxiliary rule supervision do not reduce the claims to their inputs.

full rationale

PoNG is an empirical architecture paper rather than a derivation chain. The model is trained with a joint loss L = CE(σ({ŷ_k}), y) + β BCE(ζ(b_r1), r) + γ BCE(ζ(b_r2), r), where r is the multi-hot ground-truth rule vector and β=25, γ=5 (Section 3). This rule supervision is an input training signal, not a reported prediction target: all headline results are answer-accuracy values on held-out test splits of PGM, I-RAVEN, A-I-RAVEN, I-RAVEN-Mesh, VAP, and VASR. The paper explicitly discloses the contribution of this supervision in Table 6, where removing both rule heads (β=γ=0) drops I-RAVEN accuracy from 95.9% to 79.7% and I-RAVEN-Mesh from 89.3% to 32.7%; the ablation text states this 'signifies high relevance of the auxiliary training signal.' This transparency precludes any fitted-parameter-renamed-as-prediction circularity. The self-citations to A-I-RAVEN and I-RAVEN-Mesh identify externally constructed benchmark datasets, and the citation to the sparse rule encoding is a published encoding method; none is a 'uniqueness theorem' or an input whose definition already contains PoNG's output. The use of one's own benchmarks is a benchmark-design risk, not a circular-derivation step. Any concern that baselines may not have received rule-label supervision is an experimental-fairness or correctness issue, not a case where a reported prediction is equivalent by construction to a fitted input. The central claim—that PoNG generalizes on these held-out distributions—is independently checkable from the released code and public datasets, so no circular step is identifiable.

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

The central claim rests on an architecture with several hand-chosen coefficients and a strong reliance on rule-label supervision. No new external entities are postulated.

free parameters (3)
  • β (rule-head loss weight) = 25
    Hand-chosen to balance the auxiliary rule prediction losses in the joint loss function.
  • γ (target-conditioned rule-head loss weight) = 5
    Hand-chosen; ablation shows removing either head reduces performance.
  • group count in P3/P4 = 3 for RPMs/VAP, 2 for VASR
    Chosen to match the number of rows in the matrix, a strong inductive bias.
assumptions (3)
  • domain assumption Ground-truth rule labels r are available for training on all benchmarks except VASR
    The loss function uses rule labels via Pr1 and Pr2; if baselines lack this supervision, comparisons are unfair.
  • standard math Standard i.i.d. train/validation/test splits reflect generalization
    Classical empirical ML assumption that held-out accuracy measures generalization.
  • domain assumption The benchmark datasets' abstract structures are correctly specified
    The paper relies on PGM/A-I-RAVEN rule definitions for evaluation and rule supervision.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Advancing Generalization Across a Variety of Abstract Visual Reasoning Tasks." pith.science (2026). https://pith.science/paper/P4ZY5AM6

@misc{pith2026250513391,
  author       = {Pith},
  title        = {Pith review of: Advancing Generalization Across a Variety of Abstract Visual Reasoning Tasks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/P4ZY5AM6}},
  note         = {Machine review of arXiv:2505.13391}
}
read the original abstract

The abstract visual reasoning (AVR) domain presents a diverse suite of analogy-based tasks devoted to studying model generalization. Recent years have brought dynamic progress in the field, particularly in i.i.d. scenarios, in which models are trained and evaluated on the same data distributions. Nevertheless, o.o.d. setups that assess model generalization to new test distributions remain challenging even for the most recent models. To advance generalization in AVR tasks, we present the Pathways of Normalized Group Convolution model (PoNG), a novel neural architecture that features group convolution, normalization, and a parallel design. We consider a wide set of AVR benchmarks, including Raven's Progressive Matrices and visual analogy problems with both synthetic and real-world images. The experiments demonstrate strong generalization capabilities of the proposed model, which in several settings outperforms the existing literature methods.

Figures

Figures reproduced from arXiv: 2505.13391 by the authors.

Figure 1
Figure 1. Raven’s Progressive Matrices (RPMs). (a) VAP [Hill et al., 2019] (b) VASR [Bitton et al., 2023] [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Visual analogies. importance of forming domain-independent concept repre￾sentations. Such analogy-making abilities are closely tied to fluid intelligence [Snow et al., 1984; Carpenter et al., 1990; Lake et al., 2017], a cornerstone of human cognition. Repli￾cating these capabilities in learning systems has been a long-standing goal of research in the field [Gentner, 1980; Hofstadter, 1995; French, 2002; Lovett et al… view at source ↗
Figure 3
Figure 3. A-I-RAVEN [Małkinski and Ma ´ ndziuk, 2025a ´ ]. (a) Number DistributeThree (b) Position Arithmetic [PITH_FULL_IMAGE:figures/full_fig_p002_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: I-RAVEN-Mesh [Małkinski and Ma ´ ndziuk, 2025a ´ ]. For instance, each PGM matrix [Barrett et al., 2018] has a corresponding abstract structure S = {(r, o, a) | r ∈ R, o ∈ O, a ∈ A}, where R = {progression, XOR, OR, AND, consistent union}, O = {shape, line}, and A = {s…
Figure 5
Figure 5. Figure 5: PoNG. (a) The panel encoder embeds each input image xi independently, producing hi. Context panel embeddings {hi} 8 i=1 together with the embedding of k’th answer hk are stacked and processed with the reasoner, leading to zk. (b) The pathways block, a key component of …
Figure 6
Figure 6. Figure 6: Selected examples of A/Color where PoNG succeeded and its ablation variant without TCN failed. In each case, the ablation model selected an answer differing in the held-out Color attribute. Missed rules: (a) Same set of 3 colors per row. (b) Decreasing color progressio…
Figure 7
Figure 7. Figure 7: Selected examples of A/Size where PoNG succeeded and its ablation variant without TCN failed. In each case, the ablation model selected an answer differing in the held-out Size attribute. Missed rules: (a) Decreasing size from left to right. (b) Constant size across ro…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 45 canonical work pages

  1. [1]

    Layer normalization

    Jimmy Lei Ba, Jamie Ryan Kiros, and Geoffrey E Hinton. Layer normalization. arXiv:1607.06450 , 2016

    arXiv 2016

  2. [2]

    Measuring abstract reasoning in neural networks

    David Barrett, Felix Hill, Adam Santoro, Ari Morcos, and Timothy Lillicrap. Measuring abstract reasoning in neural networks. In ICML , pages 511--520. PMLR, 2018

    work page 2018

  3. [3]

    Scale-localized abstract reasoning

    Yaniv Benny, Niv Pekar, and Lior Wolf. Scale-localized abstract reasoning. In CVPR , pages 12557--12565, 2021

    work page 2021

  4. [4]

    VASR : Visual analogies of situation recognition

    Yonatan Bitton, Ron Yosef, Eliyahu Strugo, Dafna Shahaf, Roy Schwartz, and Gabriel Stanovsky. VASR : Visual analogies of situation recognition. In AAAI , volume 37, pages 241--249, 2023

    work page 2023

  5. [5]

    What one intelligence test measures: a theoretical account of the processing in the raven progressive matrices test

    Patricia A Carpenter, Marcel A Just, and Peter Shell. What one intelligence test measures: a theoretical account of the processing in the raven progressive matrices test. Psychological review , 97(3):404, 1990

    work page 1990

  6. [6]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale. In ICLR , 2021

    work page 2021

  7. [7]

    The computational modeling of analogy-making

    Robert M French. The computational modeling of analogy-making. Trends in cognitive Sciences , 6(5):200--205, 2002

    work page 2002

  8. [8]

    Computational models of analogy

    Dedre Gentner and Kenneth D Forbus. Computational models of analogy. Wiley interdisciplinary reviews: cognitive science , 2(3):266--276, 2011

    work page 2011

Show all 50 references
  1. [9]

    The structure of analogical models in science

    Dedre Gentner. The structure of analogical models in science . Bolt Beranek and Newman Cambridge, 1980

    1980

  2. [10]

    Computer models solving intelligence test problems: Progress and implications

    Jos \'e Hern \'a ndez-Orallo, Fernando Mart \' nez-Plumed, Ute Schmid, Michael Siebers, and David L Dowe. Computer models solving intelligence test problems: Progress and implications. Artificial Intelligence , 230:74--107, 2016

    2016

  3. [11]

    Learning to make analogies by contrasting abstract relational structure

    Felix Hill, Adam Santoro, David Barrett, Ari Morcos, and Timothy Lillicrap. Learning to make analogies by contrasting abstract relational structure. In ICLR , 2019

    2019

  4. [12]

    Fluid concepts and creative analogies: Computer models of the fundamental mechanisms of thought

    Douglas R Hofstadter. Fluid concepts and creative analogies: Computer models of the fundamental mechanisms of thought . Basic books, 1995

    1995

  5. [13]

    IQ of neural networks

    Dokhyam Hoshen and Michael Werman. IQ of neural networks. arXiv:1710.01692 , 2017

    2017 arXiv

  6. [14]

    Stratified rule-aware network for abstract visual reasoning

    Sheng Hu, Yuqing Ma, Xianglong Liu, Yanlu Wei, and Shihao Bai. Stratified rule-aware network for abstract visual reasoning. In AAAI , volume 35, pages 1567--1574, 2021

    2021

  7. [15]

    Batch normalization: Accelerating deep network training by reducing internal covariate shift

    Sergey Ioffe and Christian Szegedy. Batch normalization: Accelerating deep network training by reducing internal covariate shift. In ICML , pages 448--456. PMLR, 2015

    2015

  8. [16]

    Adam: A method for stochastic optimization

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. ICLR , 2014

    2014

  9. [17]

    Imagenet classification with deep convolutional neural networks

    Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. Imagenet classification with deep convolutional neural networks. NeurIPS , 25:1097--1105, 2012

    2012

  10. [18]

    Building machines that learn and think like people

    Brenden M Lake, Tomer D Ullman, Joshua B Tenenbaum, and Samuel J Gershman. Building machines that learn and think like people. Behavioral and brain sciences , 40, 2017

    2017

  11. [19]

    Analogy with qualitative spatial representations can simulate solving raven's progressive matrices

    Andrew Lovett, Kenneth Forbus, and Jeffrey Usher. Analogy with qualitative spatial representations can simulate solving raven's progressive matrices. In Proc. of the Annual Meeting of the Cognitive Science Society , volume 29, 2007

    2007

  12. [20]

    A review of emerging research directions in abstract visual reasoning

    Miko aj Ma ki \'n ski and Jacek Ma \'n dziuk. A review of emerging research directions in abstract visual reasoning. Information Fusion , 91:713--736, 2023

    2023

  13. [21]

    Multi-label contrastive learning for abstract visual reasoning

    Miko aj Ma ki \'n ski and Jacek Ma \'n dziuk. Multi-label contrastive learning for abstract visual reasoning. IEEE Transactions on Neural Networks and Learning Systems , 35(2):1941--1953, 2024

    1941

  14. [22]

    One self-configurable model to solve many abstract visual reasoning problems

    Miko aj Ma ki \'n ski and Jacek Ma \'n dziuk. One self-configurable model to solve many abstract visual reasoning problems. In AAAI , volume 38, pages 14297--14305, 2024

    2024

  15. [23]

    A-I-RAVEN and I-RAVEN-Mesh : Two new benchmarks for abstract visual reasoning

    Miko aj Ma ki \'n ski and Jacek Ma \'n dziuk. A-I-RAVEN and I-RAVEN-Mesh : Two new benchmarks for abstract visual reasoning. arXiv:2406.11061 , 2025

    2025 arXiv

  16. [24]

    Deep learning methods for abstract visual reasoning: A survey on raven's progressive matrices

    Miko aj Ma ki \'n ski and Jacek Ma \'n dziuk. Deep learning methods for abstract visual reasoning: A survey on raven's progressive matrices. ACM Computing Surveys , 57(7):1--36, 2025

    2025

  17. [25]

    DeepIQ : A human-inspired AI system for solving IQ test problems

    Jacek Ma \'n dziuk and Adam \.Z ychowski. DeepIQ : A human-inspired AI system for solving IQ test problems. In 2019 International Joint Conference on Neural Networks , pages 1--8. IEEE, 2019

    2019

  18. [26]

    Recreating raven’s: Software for systematically generating large numbers of raven-like matrix problems with normed properties

    Laura E Matzen, Zachary O Benz, Kevin R Dixon, Jamie Posey, James K Kroger, and Ann E Speed. Recreating raven’s: Software for systematically generating large numbers of raven-like matrix problems with normed properties. Behavior research methods , 42(2):525--541, 2010

    2010

  19. [27]

    Abstraction and analogy-making in artificial intelligence

    Melanie Mitchell. Abstraction and analogy-making in artificial intelligence. Annals of the New York Academy of Sciences , 1505(1):79--101, 2021

    2021

  20. [28]

    Learning to reason over visual objects

    Shanka Subhra Mondal, Taylor Whittington Webb, and Jonathan Cohen. Learning to reason over visual objects. In ICLR , 2023

    2023

  21. [29]

    Cohen, and Taylor Whittington Webb

    Shanka Subhra Mondal, Jonathan D. Cohen, and Taylor Whittington Webb. Slot abstractors: Toward scalable abstract visual reasoning. In ICML , volume 235, pages 36088--36105. PMLR, 2024

    2024

  22. [30]

    Raven's progressive matrices and vocabulary scales

    John C Raven and John Hugh Court. Raven's progressive matrices and vocabulary scales . Oxford pyschologists Press Oxford, England, 1998

    1998

  23. [31]

    Mental tests used in genetic studies: The performance of related individuals on tests mainly educative and mainly reproductive

    James C Raven. Mental tests used in genetic studies: The performance of related individuals on tests mainly educative and mainly reproductive. Master’s thesis, University of London , 1936

    1936

  24. [32]

    A simple neural network module for relational reasoning

    Adam Santoro, David Raposo, David G Barrett, Mateusz Malinowski, Razvan Pascanu, Peter Battaglia, and Timothy Lillicrap. A simple neural network module for relational reasoning. NeurIPS , 30:4967--4976, 2017

    2017

  25. [33]

    Shashank Shekhar and Graham W. Taylor. Neural structure mapping for learning abstract visual analogies, 2022

    2022

  26. [34]

    The topography of ability and learning correlations

    Richard E Snow, Patrick C Kyllonen, and Brachia Marshalek. The topography of ability and learning correlations . Advances in the psychology of human intelligence , 2(S 47):103, 1984

    1984

  27. [35]

    A closer look at generalisation in RAVEN

    Steven Spratley, Krista Ehinger, and Tim Miller. A closer look at generalisation in RAVEN . In European Conference on Computer Vision , pages 601--616. Springer, 2020

    2020

  28. [36]

    Evaluating the progress of deep learning for visual relational concepts

    Sebastian Stabinger, David Peer, Justus Piater, and Antonio Rodr \' guez-S \'a nchez. Evaluating the progress of deep learning for visual relational concepts. Journal of Vision , 21(11):8--8, 2021

    2021

  29. [37]

    Duel-based deep learning system for solving IQ tests

    Paulina Tomaszewska, Adam \.Z ychowski, and Jacek Ma \'n dziuk. Duel-based deep learning system for solving IQ tests. In International Conference on Artificial Intelligence and Statistics , pages 10483--10492. PMLR, 2022

    2022

  30. [38]

    How much intelligence is there in artificial intelligence? a 2020 update

    Han LJ van der Maas, Lukas Snoek, and Claire E Stevenson. How much intelligence is there in artificial intelligence? a 2020 update. Intelligence , 87:101548, 2021

    2020

  31. [39]

    Automatic generation of raven’s progressive matrices

    Ke Wang and Zhendong Su. Automatic generation of raven’s progressive matrices. In Twenty-fourth international joint conference on artificial intelligence , 2015

    2015

  32. [40]

    Learning representations that support extrapolation

    Taylor Webb, Zachary Dulberg, Steven Frankland, Alexander Petrov, Randall O’Reilly, and Jonathan Cohen. Learning representations that support extrapolation. In ICML , pages 10136--10146. PMLR, 2020

    2020

  33. [41]

    The scattering compositional learner: Discovering objects, attributes, relationships in analogical reasoning

    Yuhuai Wu, Honghua Dong, Roger Grosse, and Jimmy Ba. The scattering compositional learner: Discovering objects, attributes, relationships in analogical reasoning. arXiv:2007.04212 , 2020

    2007 arXiv

  34. [42]

    Neural prediction errors enable analogical visual reasoning in human standard intelligence tests

    Lingxiao Yang, Hongzhi You, Zonglei Zhen, Dahui Wang, Xiaohong Wan, Xiaohua Xie, and Ru-Yuan Zhang. Neural prediction errors enable analogical visual reasoning in human standard intelligence tests. In ICML , volume 202, pages 39572--39583. PMLR, 2023

    2023

  35. [43]

    A cognitively-inspired neural architecture for visual abstract reasoning using contrastive perceptual and conceptual processing

    Yuan Yang, Deepayan Sanyal, James Ainooson, Joel Michelson, Effat Farhana, and Maithilee Kunda. A cognitively-inspired neural architecture for visual abstract reasoning using contrastive perceptual and conceptual processing. arXiv:2309.10532 , 2023

    2023 arXiv

  36. [44]

    RAVEN : A dataset for relational and analogical visual reasoning

    Chi Zhang, Feng Gao, Baoxiong Jia, Yixin Zhu, and Song-Chun Zhu. RAVEN : A dataset for relational and analogical visual reasoning. In CVPR , pages 5317--5327, 2019

    2019

  37. [45]

    Learning perceptual inference by contrasting

    Chi Zhang, Baoxiong Jia, Feng Gao, Yixin Zhu, Hongjing Lu, and Song-Chun Zhu. Learning perceptual inference by contrasting. NeurIPS , 32:1075--1087, 2019

    2019

  38. [46]

    Abstract spatial-temporal reasoning via probabilistic abduction and execution

    Chi Zhang, Baoxiong Jia, Song-Chun Zhu, and Yixin Zhu. Abstract spatial-temporal reasoning via probabilistic abduction and execution. In CVPR , pages 9736--9746, 2021

    2021

  39. [47]

    Learning algebraic representation for systematic generalization in abstract reasoning

    Chi Zhang, Sirui Xie, Baoxiong Jia, Ying Nian Wu, Song-Chun Zhu, and Yixin Zhu. Learning algebraic representation for systematic generalization in abstract reasoning. In European Conference on Computer Vision , pages 692--709. Springer, 2022

    2022

  40. [48]

    Learning robust rule representations for abstract reasoning via internal inferences

    Wenbo Zhang, Site Mo, Xianggen Liu, Sen Song, et al. Learning robust rule representations for abstract reasoning via internal inferences. NeurIPS , 35:33550--33562, 2022

    2022

  41. [49]

    Learning visual abstract reasoning through dual-stream networks

    Kai Zhao, Chang Xu, and Bailu Si. Learning visual abstract reasoning through dual-stream networks. In AAAI , volume 38, pages 16979--16988, 2024

    2024

  42. [50]

    write newline

    " write newline "" before.all 'output.state := FUNCTION fin.entry add.period write newline FUNCTION new.block output.state before.all = 'skip after.block 'output.state := if FUNCTION new.sentence output.state after.block = 'skip output.state before.all = 'skip after.sentence '...

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.