Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Granular Concept Circuits: Toward a Fine-Grained Circuit Discovery for Concept Representations

T0 review · 4 major / 5 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 The paper claims that a deep vision model's representation of a query image can be decomposed into multiple fine-grained concept circuits.

desk verdict A genuinely new method for finding multiple per-query concept circuits, but the claim that each circuit encodes a single distinct concept is not yet validated by the experiments. read the letter →

arxiv 2508.01728 v1 pith:MKTE3N5Y submitted 2025-08-03 cs.CV cs.AI

Dahee Kwon , Sehyun Lee , Jaesik Choi This is my paper
classification cs.CVcs.AI
keywords circuitdiscoveryconceptinterpretabilityneuronablationsemanticalignmentvisualconceptsmechanisticdeepvisionmodelsfine-grainedexplanation
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 tries to establish that a deep image classifier's distributed representation of a query image can be decomposed into multiple granular circuits, each encoding one human-interpretable visual concept relevant to that query. It argues that prior circuit methods either produce a single class-level pathway or miss concept specificity, whereas the proposed method automatically returns several circuits per query. If true, this gives a fine-grained operational account of where and how concepts such as 'sky background', 'flags', or 'clock' live inside a trained model, and supports uses like auditing misclassifications and finding concepts shared across different classes. The central evidence is that ablating neurons inside the discovered circuits degrades model outputs far more than ablating the same number of neurons outside them.

What carries the argument

The mechanism is a forward, iterative connectivity search through the network. Starting from root nodes, defined as neurons whose activations rank in the top 1 percent for the query, the method scores each candidate edge from a source neuron in layer $l$ to a target neuron in layer $l+1$ with two quantities: the Neuron Sensitivity Score, which zero-masks the source neuron and measures the resulting change in the target's activation, and the Semantic Flow Score, which computes the overlap of the top-$k$ highly activated samples between the two neurons. Edges pass only if both scores clear thresholds, with the Neuron Sensitivity threshold selected automatically by peak-over-threshold extreme-value statistics and the Semantic Flow threshold set to the average score. Surviving edges are added to the circuit, the newly added nodes become new sources, and the search continues until no further connections remain; repeating this for every root node yields the collection of granular concept circuits.

What would settle it

Build a small network with a planted concept circuit in which the next-layer response depends on the joint activity of two source neurons rather than on either alone, run GCC on it, and check whether both edges are recovered; a failure to connect either neuron would show that the first-order zero-masking proxy misses higher-order concept dependencies.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is that each discovered Granular Concept Circuit corresponds to a coherent visual concept relevant to the given query, and that the full set of circuits forms a fine-grained decomposition of the model's representation for that query. The claim includes a methodological assertion: connections between adjacent-layer neurons that carry a single concept can be identified by combining a first-order functional-dependence measure, the Neuron Sensitivity Score, with a semantic-alignment measure, the Semantic Flow Score. The paper further claims it is the first circuit discovery method to return multiple concept-specific circuits rather than a single class-level circuit, and it validates this on CNNs and transformer models with ablation experiments and a user study.

Load-bearing premise

The method assumes that a first-order ablation of one neuron at a time, combined with overlap of the most strongly activated samples, is enough to identify connections that carry a single coherent concept across layers; if higher-order interactions or polysemantic neurons break this proxy, the discovered circuits may be statistically sensitive pathways without stable semantic meaning.

Editorial extensions

If this is right

  • Ablating the neurons of a discovered circuit produces a substantially larger drop in the model's output than ablating the same number of neurons outside it, which the paper reports as evidence that the circuits are both faithful and complete.
  • Because each circuit is tied to a specific concept rather than a class, the method separates multiple concepts within one query image and shows how each concept flows through successive layers.
  • For multiple queries, the method can expose concepts shared across different classes, such as a common 'wheel' pattern among vehicles, by propagating forward from root nodes instead of relying on class logits.
  • In a misclassification audit, inhibiting or stimulating the neurons of discovered circuits changes the logit for the true class, revealing which concept-level pathways drive the erroneous prediction.
  • The same circuit-ablation procedure applied to transformer models yields large accuracy drops, indicating that the discovery mechanism transfers beyond convolutional architectures.

Reading between the lines

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

  • Beyond the paper's claims, the first-order ablation proxy could be stress-tested on synthetic networks with planted concept circuits: if a concept is encoded by two neurons whose joint effect, but neither alone, drives the next layer, zero-masking one source at a time may fail to find the edge.
  • An extension not tested in the paper would replace the sample-overlap Semantic Flow Score with a concept-identity metric, such as clustering or textual descriptions of neuron activations, and check whether the discovered circuits remain stable under that change.
  • The paper itself notes that a single concept may be distributed across multiple circuit pathways under strict connectivity thresholds, so the number of circuits per query should be treated as a tunable granularity rather than a fixed ground-truth decomposition.
  • Because the forward search does not require class labels, the method could be redirected at attributes, textures, or background objects; one could ask which circuits fire for a specific property and audit the model's reasoning about that property.
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

4 major / 5 minor

Summary. The paper introduces Granular Concept Circuits (GCC), a method for discovering sets of neurons across layers in vision models that are claimed to each encode a distinct, human-interpretable visual concept relevant to a given query image. Connectivity between adjacent-layer neurons is established through two scores: a Neuron Sensitivity Score (zero-masking the source neuron and measuring next-layer activation change, Eq. 1) and a Semantic Flow Score (overlap of top-k highly activated samples, Eq. 3). Starting from root nodes selected by high query activation, the method iteratively grows each circuit and repeats for all root nodes, producing multiple concept-specific circuits. Experiments on CNNs (VGG19, ResNet50/101, MobileNetV3) and Transformers (ViT, Swin, CLIP-ViT) include qualitative visualizations, logit/accuracy ablation comparisons, edge deletion/insertion curves, a 33-participant user study, and two use cases (misclassification audit and common-concept discovery). The paper claims to be the first to identify fine-grained, concept-specific visual circuits.

Significance. If the central claim holds, GCC would be a valuable contribution to vision interpretability by moving from single-neuron or single-circuit analyses to a fine-grained, concept-wise decomposition of a model's internal representations. The paper has several strengths: the code is publicly available; the method is evaluated across a wide range of architectures (CNNs and Transformers); the Peak-over-Threshold procedure for threshold selection is principled and empirically examined; and the user study, while small, is a positive step toward human validation. However, the key claim that each discovered circuit is a single, coherent human-interpretable concept is not yet convincingly demonstrated. The quantitative evaluations validate aggregate influence rather than per-circuit concept specificity, and the faithfulness metric is partly self-confirming because it ablates exactly the neurons that the method selected by their influence. The first-order approximation underlying the Neuron Sensitivity Score is load-bearing for the notion of a 'concept-preserving' circuit and remains untested against alternative interaction-aware connectivity measures.

major comments (4)
  1. [Sec. 3.1, Eq. (1)] The first-order zero-masking approximation is load-bearing for the paper's central claim, but the manuscript does not validate that it preserves concept identity. If a concept is encoded by a set of neurons whose joint effect is necessary, zero-masking any single source neuron may produce only a small change in the target, so S_NS would miss the true connection. The paper acknowledges the intractability of full interactions (Sec. 3.1) and later concedes that 'a single concept may be distributed across multiple circuit pathways' (Sec. 5). These two statements together identify a concrete risk: the discovered circuits may be statistically sensitive pathways without coherent semantic meaning. I recommend adding a targeted experiment, for example comparing S_NS against second-order or joint ablations on a small subset of layers, to show that first-order effects are sufficient to recover concept-preserving connections.
  2. [Sec. 4.2, Table 2 and Appendix B] The quantitative faithfulness/completeness evaluation ablates all circuit neurons simultaneously, so it validates the union of circuits, not the specificity of each individual circuit. The paper's central claim is that 'multiple circuits, each encoding a distinct concept relevant to the query' (Abstract, Sec. 3.2). Table 2 and Table E report aggregate logit/accuracy drops when all discovered circuits are ablated together; such results would also be observed if the method found one large, unspecific influential pathway. To support the per-circuit claim, the authors should evaluate each circuit separately, for example by ablating a single circuit and showing a selective effect on concept-related outputs or by demonstrating that circuits are not redundant with one another. In addition, the average logit drops in Table 2 and the accuracy drops in Table E are reported as point estimates without error bars or significance tests over the 100 queries; at minimum, standard errors or confidence intervals should be provided.
  3. [Sec. 4.2, Table 2] The main faithfulness metric is partially self-confirming. Root nodes are selected by high query activation, edges are filtered by high S_NS and S_SF, and then the evaluation ablates these same selected neurons and measures the resulting logit drop. Any criterion that selects high-influence neurons would produce some logit drop, so the large drop in Table 2 does not distinguish concept-specific circuits from a generic set of influential neurons. The random baseline is not matched on activation magnitude or on query relevance, and the complement ablation (OursC) excludes by construction the very neurons the method deems important. A stronger control would ablate the same number of top-activation neurons (without the connectivity constraints) or a set matched on S_NS but not on S_SF, to show that the joint connectivity and semantic alignment criteria are what matter.
  4. [Sec. 4.3 and Sec. 4.1] The claim that each GCC is a single human-interpretable concept rests primarily on qualitative examples and a 33-participant user study, but the user study does not directly test monosemanticity. The questions in Fig. 5 ask about query relevance, diversity, and prototypicality of the circuit as a whole, not whether each circuit corresponds to one coherent concept that a human can label consistently. The qualitative visualizations in Figs. 3 and 4 are suggestive, but they show only a few examples and the manual textual descriptions could be post-hoc rationalizations. I recommend a more objective concept-coherence metric, such as NetDissect-style IoU against semantic segmentation masks for the concept named by the circuit, or a label-consistency measure across the top highly activated samples of the circuit's root and downstream nodes.
minor comments (5)
  1. [Sec. 3.1, Eq. (1)] The notation f^{l+1}(a^l_c) is ambiguous: it appears to feed the masked or unmasked activation vector of the entire layer, but the subscript c suggests a single channel. Please clarify that a^l_c denotes the full layer activation with only neuron c zero-masked, so that the expression reads as a function of the whole vector.
  2. [Sec. 4.2, Fig. 6] The paper refers to 'A VCD' in the text near Fig. 6; this appears to be a typo for ADVC (from Ref. [29]). Please correct for consistency.
  3. [Sec. 4, Visualization Method] The phrase 'to supporess weak activations' contains a typo ('suppress'). Also, the visualization method description would benefit from stating how the threshold for the binary mask is chosen, since this affects the perceived concept quality in Figs. 3 and 4.
  4. [Appendix A.2] The root-node selection threshold (top 1%) is stated to be tunable, and the appendix mentions that top 10% increases diversity but reduces query relevance, yet no quantitative comparison is given. A small sensitivity analysis, even on one model, would help readers understand the impact of this parameter on the faithfulness/completeness metrics.
  5. [Table 3 / Table D] Table D in the appendix reports POT threshold comparisons but appears to have a missing entry for the 'Ours' row at POT=70; the table formatting should be checked. Also, the caption labels the table as 'Table D' while the text in Appendix A.3 refers to it; ensure all cross-references are consistent.

Circularity Check

2 steps flagged · score 6.0 of 10

Semantic coherence is enforced by top-k overlap and faithfulness is measured by the same zero-masking operation used to select circuits; the central concept claim rests on construction, not independent evidence.

  1. self definitional [Section 3.1, Eq. (3); Section 4, Visualization Method]
    "SSF = |Ssrc ∩ Stgt| / |Ssrc| (3) where Ssrc denotes the set of highly activated samples for the source node, and Stgt represents the corresponding set for the target node. ... The concept associated with a neuron is inferred from its highly activated samples, with four representative images per node selected from the top 10 highly activated samples in the validation set."

    The connectivity criterion S_SF is literally the overlap of top-k highly activated samples. Every edge that survives the threshold therefore shares many of the same top-activation samples by construction. When circuits are later visualized using those same top-10 activated cropped images, the apparent semantic alignment of connected nodes is a direct display of the input filter, not an independent discovery. The paper presents this enforced overlap as evidence that each circuit represents a concept, but the coherence along edges is true by definition.

  2. other [Section 3.1, Eq. (1); Section 4.2, Quantitative Results]
    "we propose a Neuron Sensitivity Score S_NS, which quantifies the influence of a source neuron on a target neuron built on an intervention-based approach by measuring the change in the target's activation when the source neuron is muted—i.e., set to zero. ... we first evaluate the effectiveness of granular concept circuits by ablating all neurons within each identified circuit and measuring the resulting change in logit outputs. A substantial drop indicates that the circuits faithfully encode query-relevant information."

    The same zero-masking operation used to select edges (S_NS) is aggregated in the evaluation: ablate circuit neurons, measure logit drop. Because edges were chosen precisely for large next-layer activation changes under masking, a large final-logit drop is statistically expected. This is a self-confirming test: it verifies that the selected neurons are intervention-important, not that they encode a distinct, human-interpretable concept. The quantitative faithfulness claim reduces to the selection criterion.

full rationale

The GCC construction itself is a concrete algorithm; the first-order sensitivity and top-k overlap are stated assumptions, not hidden fits. The circularity is in the validation logic: (1) edges are defined by S_SF top-k overlap, so the displayed semantic coherence of connected nodes is true by construction; (2) the faithfulness experiment ablates circuit neurons and measures logit drop, which is the same zero-masking intervention used to compute S_NS, so the large drop is a restatement of the selection criterion rather than independent confirmation. No load-bearing self-citation was found; reference [6] by the current authors is only listed among neuron-based interpretability methods. The central claim that each circuit encodes a distinct human-interpretable concept therefore lacks an independent test: the quantitative metric is self-confirming and the user study is subjective. Score 6 reflects partial circularity, not complete equivalence.

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

The method rests on several domain assumptions about how concepts are encoded in neurons (activation-pattern overlap, first-order causal sensitivity, adjacent-layer DAG structure). These are reasonable heuristics but are not proven; they are the price of the method's tractability.

free parameters (4)
  • top-k for Semantic Flow Score = unspecified (visualizations use top-10)
    S_SF requires selecting k highly activated samples per neuron; the value of k is never stated in the method section, making the semantic-alignment filter underspecified.
  • root node activation percentile = top 1% (tunable, e.g., top 10%)
    Root nodes are defined as neurons whose activation rank is within the top 1% across the dataset; the paper notes this can be adjusted for diversity, so it is a hand-chosen threshold.
  • POT threshold percentile for tau_NS = recommended 90-95
    tau_NS is set via the Peak-over-Threshold method; the paper tests 60-95 and recommends 90-95 based on empirical consistency, making it a tuned hyperparameter.
  • tau_SF as average S_SF = layer-wise average
    tau_SF is defined as the average S_SF of all nodes in the target layer; this adaptive threshold is a design choice rather than a derived value.
assumptions (4)
  • domain assumption A neuron's encoded information is reflected in the set of samples where it is highly activated.
    S_SF assumes that overlap of top-k activating samples between two neurons indicates semantic alignment. This is stated in Section 3.1 as motivation but not validated.
  • domain assumption First-order single-neuron ablation is sufficient to capture functional dependencies between neurons.
    Equation (1) measures the change in the next layer when a single source neuron is zero-masked, ignoring higher-order interactions that the paper acknowledges (O(2^|N|) intractability), so the approximation is load-bearing.
  • domain assumption Conceptually meaningful connections only occur between adjacent layers and form a DAG.
    The iterative search connects each node only to nodes in the immediately next layer (Algorithm 1); residual/skip connections are not traced as direct edges, so cross-layer pathways are omitted.
  • domain assumption Highly activated neurons carry significant query-relevant information.
    Root node selection (top 1% activated neurons) is justified by prior studies, but the specific threshold is an assumption about where concepts arise.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Granular Concept Circuits: Toward a Fine-Grained Circuit Discovery for Concept Representations." pith.science (2026). https://pith.science/paper/MKTE3N5Y

@misc{pith2026250801728,
  author       = {Pith},
  title        = {Pith review of: Granular Concept Circuits: Toward a Fine-Grained Circuit Discovery for Concept Representations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MKTE3N5Y}},
  note         = {Machine review of arXiv:2508.01728}
}
read the original abstract

Deep vision models have achieved remarkable classification performance by leveraging a hierarchical architecture in which human-interpretable concepts emerge through the composition of individual neurons across layers. Given the distributed nature of representations, pinpointing where specific visual concepts are encoded within a model remains a crucial yet challenging task. In this paper, we introduce an effective circuit discovery method, called Granular Concept Circuit (GCC), in which each circuit represents a concept relevant to a given query. To construct each circuit, our method iteratively assesses inter-neuron connectivity, focusing on both functional dependencies and semantic alignment. By automatically discovering multiple circuits, each capturing specific concepts within that query, our approach offers a profound, concept-wise interpretation of models and is the first to identify circuits tied to specific visual concepts at a fine-grained level. We validate the versatility and effectiveness of GCCs across various deep image classification models.

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Latent Anomaly Knowledge Excavation: Unveiling Sparse Sensitive Neurons in Vision-Language Models

    cs.CV 2026-04 unverdicted novelty 6.0 of 10

    LAKE identifies sparse anomaly-sensitive neurons in pre-trained VLMs using minimal normal samples to build compact normality representations and achieve SOTA anomaly detection with neuron-level interpretability.

Reference graph

Works this paper leans on

42 extracted references · 32 canonical work pages · cited by 1 Pith paper

  1. [1]

    From attribution maps to human-understandable explanations through concept rele- vance propagation

    Reduan Achtibat, Maximilian Dreyer, Ilona Eisenbraun, Sebastian Bosse, Thomas Wiegand, Wojciech Samek, and Sebastian Lapuschkin. From attribution maps to human-understandable explanations through concept rele- vance propagation. Nature Machine Intelligence, 2023. 2, 3, 4

    work page 2023

  2. [2]

    Cub-200-resnet50 pretrained model,

    Anonymous Authors. Cub-200-resnet50 pretrained model,

    work page

  3. [3]

    Network dissection: Quantifying inter- pretability of deep visual representations

    David Bau, Bolei Zhou, Aditya Khosla, Aude Oliva, and Antonio Torralba. Network dissection: Quantifying inter- pretability of deep visual representations. In Proceedings of the IEEE conference on computer vision and pattern recog- nition, 2017. 1, 2, 4

    work page 2017

  4. [4]

    Understanding the role of individual units in a deep neural network.Proceedings of the National Academy of Sciences, 2020

    David Bau, Jun-Yan Zhu, Hendrik Strobelt, Agata Lapedriza, Bolei Zhou, and Antonio Torralba. Understanding the role of individual units in a deep neural network.Proceedings of the National Academy of Sciences, 2020. 1, 2

    work page 2020

  5. [5]

    Rewiring the connectome: evidence and effects

    Sophie H Bennett, Alastair J Kirby, and Gerald T Finnerty. Rewiring the connectome: evidence and effects. Neuro- science & Biobehavioral Reviews, 2018. 2

    work page 2018

  6. [6]

    Under- standing distributed representations of concepts in deep neu- ral networks without supervision

    Wonjoon Chang, Dahee Kwon, and Jaesik Choi. Under- standing distributed representations of concepts in deep neu- ral networks without supervision. In Proceedings of the AAAI Conference on Artificial Intelligence, 2024. 2

    work page 2024

  7. [7]

    Towards automated circuit discovery for mechanistic interpretability

    Arthur Conmy, Augustine Mavor-Parker, Aengus Lynch, Stefan Heimersheim, and Adri `a Garriga-Alonso. Towards automated circuit discovery for mechanistic interpretability. Advances in Neural Information Processing Systems , 2023. 3

    work page 2023

  8. [8]

    Knowledge neurons in pretrained transform- ers

    Damai Dai, Li Dong, Yaru Hao, Zhifang Sui, Baobao Chang, and Furu Wei. Knowledge neurons in pretrained transform- ers. arXiv preprint arXiv:2104.08696, 2021. 4

    arXiv 2021

Show all 42 references
  1. [9]

    Models for ex- ceedances over high thresholds

    Anthony C Davison and Richard L Smith. Models for ex- ceedances over high thresholds. Journal of the Royal Statis- tical Society Series B: Statistical Methodology, 1990. 5

    1990

  2. [10]

    From colors to classes: Emergence of concepts in vision trans- formers

    Teresa Dorszewski, Lenka T ˇetkov´a, Robert Jenssen, Lars Kai Hansen, and Kristoffer Knutsen Wickstrøm. From colors to classes: Emergence of concepts in vision trans- formers. arXiv preprint arXiv:2503.24071, 2025. 1

    2025 arXiv

  3. [11]

    Craft: Concept recursive activation factoriza- tion for explainability

    Thomas Fel, Agustin Picard, Louis Bethune, Thibaut Boissin, David Vigouroux, Julien Colin, R ´emi Cad`ene, and Thomas Serre. Craft: Concept recursive activation factoriza- tion for explainability. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recogn...

    2023

  4. [12]

    Transformer feed-forward layers are key-value memories

    Mor Geva, Roei Schuster, Jonathan Berant, and Omer Levy. Transformer feed-forward layers are key-value memories. arXiv preprint arXiv:2012.14913, 2020. 1

    2012 arXiv

  5. [13]

    Towards automatic concept-based explanations

    Amirata Ghorbani, James Wexler, James Y Zou, and Been Kim. Towards automatic concept-based explanations. Ad- vances in neural information processing systems, 2019. 2

    2019

  6. [14]

    Extreme value theory: an introduction

    Laurens Haan and Ana Ferreira. Extreme value theory: an introduction. Springer, 2006. 5

    2006

  7. [15]

    The organization of behavior: A neu- ropsychological theory

    Donald Olding Hebb. The organization of behavior: A neu- ropsychological theory. Psychology press, 2005. 2

    2005

  8. [16]

    Natural language descriptions of deep visual features

    Evan Hernandez, Sarah Schwettmann, David Bau, Teona Bagashvili, Antonio Torralba, and Jacob Andreas. Natural language descriptions of deep visual features. In Interna- tional Conference on Learning Representations, 2021. 2

    2021

  9. [17]

    Distributed representations

    Geoffrey E Hinton. Distributed representations. 1984. 1

    1984

  10. [18]

    Learning distributed representations of concepts

    Geoffrey E Hinton. Learning distributed representations of concepts. In Proceedings of the Annual Meeting of the Cog- nitive Science Society, 1986. 1

    1986

  11. [19]

    Synaptic plasticity-dependent competition rule influences memory formation

    Yire Jeong, Hye-Yeon Cho, Mujun Kim, Jung-Pyo Oh, Min Soo Kang, Miran Yoo, Han-Sol Lee, and Jin-Hee Han. Synaptic plasticity-dependent competition rule influences memory formation. Nature communications, 2021. 2

    2021

  12. [20]

    Identifying inter- pretable subspaces in image representations

    Neha Kalibhat, Shweta Bhardwaj, C Bayan Bruss, Hamed Firooz, Maziar Sanjabi, and Soheil Feizi. Identifying inter- pretable subspaces in image representations. InInternational Conference on Machine Learning. PMLR, 2023. 2, 4

    2023

  13. [21]

    Interpretability be- yond feature attribution: Quantitative testing with concept activation vectors (tcav)

    Been Kim, Martin Wattenberg, Justin Gilmer, Carrie Cai, James Wexler, Fernanda Viegas, et al. Interpretability be- yond feature attribution: Quantitative testing with concept activation vectors (tcav). In International conference on ma- chine learning, pages 2668–2677. PMLR, 2018. 2

    2018

  14. [22]

    Visual concept connectome (vcc): Open world con- cept discovery and their interlayer connections in deep mod- els

    Matthew Kowal, Richard P Wildes, and Konstantinos G Der- panis. Visual concept connectome (vcc): Open world con- cept discovery and their interlayer connections in deep mod- els. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition , pages 108...

  15. [23]

    Sparse feature cir- cuits: Discovering and editing interpretable causal graphs in language models

    Samuel Marks, Can Rager, Eric J Michaud, Yonatan Be- linkov, David Bau, and Aaron Mueller. Sparse feature cir- cuits: Discovering and editing interpretable causal graphs in language models. arXiv preprint arXiv:2403.19647, 2024. 3

    2024 arXiv

  16. [24]

    Locating and editing factual associations in gpt

    Kevin Meng, David Bau, Alex Andonian, and Yonatan Be- linkov. Locating and editing factual associations in gpt. Ad- vances in neural information processing systems, 2022. 3

    2022

  17. [25]

    Oikarinen and Tsui-Wei Weng

    Tuomas P. Oikarinen and Tsui-Wei Weng. Clip-dissect: Au- tomatic description of neuron representations in deep vi- sion networks. In The Eleventh International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1-5, 2023. OpenReview.net, 2023. 2

    2023

  18. [26]

    An overview of early vision in inceptionv1

    Chris Olah, Nick Cammarata, Ludwig Schubert, Gabriel Goh, Michael Petrov, and Shan Carter. An overview of early vision in inceptionv1. Distill, 2020. 1, 2, 3

    2020

  19. [27]

    Rise: Random- ized input sampling for explanation of black-box models

    Vitali Petsiuk, Abir Das, and Kate Saenko. Rise: Random- ized input sampling for explanation of black-box models. arXiv preprint arXiv:1806.07421, 2018. 7

    2018 arXiv

  20. [28]

    Statistical inference using extreme order statistics

    James Pickands III. Statistical inference using extreme order statistics. the Annals of Statistics, 1975. 5

    1975

  21. [29]

    Automatic discovery of visual circuits

    Achyuta Rajaram, Neil Chowdhury, Antonio Torralba, Jacob Andreas, and Sarah Schwettmann. Automatic discovery of visual circuits. arXiv preprint arXiv:2404.14349, 2024. 2, 3, 7

    2024 arXiv

  22. [30]

    Distributed represen- tations in memory: insights from functional brain imaging

    Jesse Rissman and Anthony D Wagner. Distributed represen- tations in memory: insights from functional brain imaging. Annual review of psychology, 2012. 2

    2012

  23. [31]

    Parallel distributed processing, volume 1: Ex- plorations in the microstructure of cognition: Foundations

    David E Rumelhart, James L McClelland, PDP Research Group, et al. Parallel distributed processing, volume 1: Ex- plorations in the microstructure of cognition: Foundations . The MIT press, 1986. 2

    1986

  24. [32]

    Genetic mecha- nisms underlying the evolution of connectivity in the human cortex

    Ewoud RE Schmidt and Franck Polleux. Genetic mecha- nisms underlying the evolution of connectivity in the human cortex. Frontiers in Neural Circuits, 2022. 2

    2022

  25. [33]

    The sankey diagram in energy and material flow management: part ii: methodology and current applica- tions

    Mario Schmidt. The sankey diagram in energy and material flow management: part ii: methodology and current applica- tions. Journal of industrial ecology, 2008. 6

    2008

  26. [34]

    Exploratory data analysis

    John Wilder Tukey et al. Exploratory data analysis . Springer, 1977. 2

    1977

  27. [35]

    The caltech-ucsd birds-200-2011 dataset

    Catherine Wah, Steve Branson, Peter Welinder, Pietro Per- ona, and Serge Belongie. The caltech-ucsd birds-200-2011 dataset. 2011. 6

    2011

  28. [36]

    Hint: Hierarchical neuron concept explainer

    Andong Wang, Wei-Ning Lee, and Xiaojuan Qi. Hint: Hierarchical neuron concept explainer. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022. 2

    2022

  29. [37]

    Interpretability in the wild: a circuit for indirect object identification in gpt-2 small

    Kevin Wang, Alexandre Variengien, Arthur Conmy, Buck Shlegeris, and Jacob Steinhardt. Interpretability in the wild: a circuit for indirect object identification in gpt-2 small. arXiv preprint arXiv:2211.00593, 2022. 3, 6

    2022 arXiv

  30. [38]

    Discovering in- fluential neuron path in vision transformers

    Yifan Wang, Yifei Liu, Yingdong Shi, Changming Li, Anqi Pang, Sibei Yang, Jingyi Yu, and Kan Ren. Discovering in- fluential neuron path in vision transformers. arXiv preprint arXiv:2503.09046, 2025. 1

    2025 arXiv

  31. [39]

    Neuroscience: Exploring the brain: By mark f

    Isaac K Wood. Neuroscience: Exploring the brain: By mark f. bear, barry w. connors and michael a. paradiso. baltimore: Williams & wilkins, 1996. pp. 666, 1996. 2

    1996

  32. [40]

    What part of the neural network does this? understanding lstms by measuring and dissecting neurons

    Ji Xin, Jimmy Lin, and Yaoliang Yu. What part of the neural network does this? understanding lstms by measuring and dissecting neurons. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natur...

    2019

  33. [41]

    Functional faithfulness in the wild: Circuit discovery with differentiable computation graph pruning

    Lei Yu, Jingcheng Niu, Zining Zhu, and Gerald Penn. Functional faithfulness in the wild: Circuit discovery with differentiable computation graph pruning. arXiv preprint arXiv:2407.03779, 2024. 3, 6

    2024

  34. [42]

    Revisiting the importance of individual units in cnns via ab- lation

    Bolei Zhou, Yiyou Sun, David Bau, and Antonio Torralba. Revisiting the importance of individual units in cnns via ab- lation. arXiv preprint arXiv:1806.02891, 2018. 4 Granular Concept Circuits: Toward a Fine-Grained Circuit Discovery for Concept Representations Supplementary M...

    2018 arXiv

Pith tools

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