Pith. sign in

REVIEW 5 major objections 4 minor 46 references

FrED: External Data Influence Estimation via Domain Knowledge Graph Grounding

T0 review · 5 major / 4 minor · reviewed 2026-08-02 · deepseek-v4-flash

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

Pith's one-line read A black-box estimator that fuses latent similarity with domain-graph evidence nearly matches gradient-based training-data attribution, without touching model internals.

desk verdict Strong black-box LDS numbers driven by a sharpened latent baseline; the KG boost is small, under-specified, and the central 'grounding' claim is not yet established. read the letter →

arxiv 2607.21615 v1 pith:YLOS74UV submitted 2026-05-26 cs.AI cs.LG

Theodoros Aivalis , Iraklis A. Klampanos , Antonis Troumpoukis , Joemon M. Jose This is my paper
classification cs.AIcs.LG
keywords trainingdataattributionblack-boxinfluenceestimationknowledgegraphsdomaingroundinggenerativediffusionmodelslineardatamodelingscoreanalogretrievalweatherforecasting
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 training-data influence can be estimated from outside a model, without accessing weights or gradients, by combining continuous feature similarity with a discrete domain knowledge graph. The method scores each training sample by a sharpened latent similarity to the target output, then multiplies that score by an asymmetric boost derived from the sample's rank in a curated graph of artistic or environmental relations. On an art-generation benchmark, this reaches a linear datamodeling score of 29.1% on validation, where a standard black-box similarity baseline scores 8.9% and a gradient-based estimator scores 27.6%, narrowing most of the gap to parametric methods. The same pipeline, applied to flood forecasts, improves geographic precision of retrieved historical analogs from 50.9% to 65.7%. If the claims hold, closed or proprietary generative models can be audited post-hoc for influence and provenance.

What carries the argument

The central object is the asymmetric boost: a sample's final attribution score is S_lat × (1 + W / (k + R_dom)), where S_lat is the latent-engine probability (feature similarity raised to a high power times a distinctiveness prior) and R_dom is the sample's integer rank in a domain knowledge graph. The graph rank is obtained by a contrastively learned projection that maps visual features onto graph-embedding coordinates, and the boost uses reciprocal-rank decay so that graph evidence re-ranks only top latent candidates while never discarding samples with missing metadata. This lets the continuous representation carry the attribution while the discrete graph supplies historical or physical co

What would settle it

Permute the node labels of the domain knowledge graph while keeping its structure and the latent scores identical, and rerun the method on the same benchmark; if the linear datamodeling score does not drop substantially from the reported 29%, the domain boost is not carrying structural information. Alternatively, hold out a set of generated images with known style lineage and check whether the learned projection's domain ranking of those images agrees with ground-truth artist and style annotations.

Watch

Extended reading notes

Core claim

FrED's central claim is that an attribution weight can be formalized as a posterior probability whose ranking is governed by the product of a latent-space likelihood and an informational-scarcity prior, and that this continuous score should be the primary evidence while a knowledge-graph rank acts only as a contextual boost. The reported art-benchmark results show this fusion substantially outperforms all black-box similarity baselines and approaches gradient-based estimators that require internal access: on the generation split the paper reports 21.3% versus 26.5% for a leading gradient-based method, and on validation 29.1% versus 27.6%, while a stronger parametric method reaches 38%. In a

Load-bearing premise

The load-bearing premise is that the contrastively learned mapping from images to graph coordinates, trained on a small unstated set of metadata-rich anchor paintings, reliably places unseen and synthetic images near their correct art-historical nodes; if that alignment fails, the domain rank in the boost is an arbitrary learned projection and the reported gains from knowledge-graph grounding collapse.

Editorial extensions

If this is right

  • Closed or proprietary generative models can be audited post-hoc for which training samples most plausibly shaped a given output, without requiring weight or gradient access.
  • Attribution responses can carry an interpretable historical or physical explanation, not just a similarity vector, because the domain graph supplies structural context.
  • The asymmetric fusion is robust to incomplete knowledge graphs: samples absent from the graph still receive their full latent score, so attribution never breaks due to missing metadata.
  • In weather forecasting, the same pipeline can retrieve physically consistent historical analogs, improving geographic localization over latent-only retrieval for decision support.
  • The reported art-benchmark numbers suggest black-box attribution can narrow the gap to parametric estimators on stylized image generation, opening a practical route to transparency for large deployed models.

Reading between the lines

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

  • The fairness of the comparison rests on the unstated size, sampling method, and coverage of the anchor subset used to train the visual-to-graph projection; if that subset is small or cherry-picked, the reported gains may partly reflect the projection's fit to the benchmark rather than genuine structural grounding. A direct test would vary the anchor subset and re-measure LDS.
  • The ablation shows that raising the similarity score to a high power, p(c|x)^10, is the largest single driver of accuracy; this hints that the method's gain may be substantially a calibration effect concentrating influence, and the knowledge graph contributes a smaller refinement. If so, the method's advantage might generalize only when influence in generative models is genuinely sparse.
  • The same asymmetric-boost fusion could extend to text-to-image provenance and retrieval-augmented generation, where domain graphs already exist or can be induced automatically; the main barrier is the 'expert bottleneck' of knowledge-graph construction, not the scoring mechanism.
  • The environmental case study stops short of causal influence; if a counterfactual evaluation protocol for weather models becomes feasible, the analog-retrieval signal could be upgraded into a true training-data attribution claim, which would be a much stronger validation of the framework.
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

5 major / 4 minor

Summary. FrED is a black-box training-data attribution method. It fuses a continuous latent-space score (ViT-g-14 cosine similarity, raised to a power p, multiplied by a distinctiveness prior) with a discrete domain-Knowledge-Graph rank obtained via a learned contrastive projection from images to graph embeddings. The two signals are combined by an asymmetric rank boost: α_j = S_lat,j × (1 + W/(k + R_dom,j)). On ArtBench, FrED is evaluated under the D-TRAK protocol using the Linear Datamodeling Score; the headline result is 29.1% on ArtBench-2 validation and 21.3% on generated targets, compared with 8.9% for CLIP cosine and 27.61%/26.53% for D-TRAK. A second, smaller study uses the same scoring idea for historical analog retrieval in flood forecasting. The paper claims that grounding latent similarity in domain KG structure is what makes the method effective, while stressing the black-box advantage over gradient-based estimators.

Significance. If the empirical claims hold, FrED would be a practically valuable black-box attribution tool: it uses only model outputs and external metadata, follows the established D-TRAK protocol, shares the exact target artifacts with the baselines, and provides an open-source code link. The paper also includes useful ablation tables and a cross-domain demonstration, which are strengths. However, the central contribution — that knowledge-graph grounding is responsible for the reported gains — is not established by the current evidence. The visual-only variant already reaches 28.5 LDS, and the full KG-boosted method reaches 29.1, a difference well within the reported confidence intervals. In addition, the boost hyperparameters W and k are never reported, and the contrastive projection that produces the domain rank is under-specified. The Bayesian framing is also explicitly a heuristic surrogate. These issues are load-bearing for the paper's main claim and require careful revision.

major comments (5)
  1. [§3.4, Eq. (5)-(6)] The fusion rule is central to FrED, but W and k in Eq. (6) are never reported anywhere in the manuscript, and S_lat,j is not defined as a formula. A reader cannot reproduce the reported 29.1 LDS or independently check whether the boost term is meaningful. Please report the exact values of W, k, and the full definition of S_lat,j, together with the procedure used to select them.
  2. [Table 7, §4.1.3] The ablation does not support the claim that KG grounding is the source of the improvement. The visual-only pipeline p(c|xj)^10 p(xj) reaches 28.5 ± 0.60 on ArtBench-2 validation, while the proposed Asymmetric Rank Boost reaches 29.1 ± 0.65 — a difference of 0.6, smaller than either confidence interval. The same pattern holds for generation (20.8 vs. 21.3). The large jump from CLIP cosine (8.9) to 28.5 is caused by the power scaling and the distinctiveness prior, not by Eq. (6). The paper should either report a significance test on the KG increment or substantially temper the claim that grounding in the domain KG drives the method.
  3. [§4.1.1, Appendix B] The contrastive visual-to-graph projection is under-specified. The text says a 'representative subset of training images with rich metadata' is used as ground-truth anchors, but the subset size, sampling method, metadata coverage, and train/validation split are not reported. More importantly, generalization to generated (out-of-distribution) images is not evaluated. If the projection is unreliable for synthetic images, the domain rank R_dom,j is an arbitrary learned quantity and the small boost from Eq. (6) carries no evidential weight. Please report these details and add a quantitative OOD validation.
  4. [§3.2, Eqs. (2)-(4)] The Bayesian derivation is presented as a probabilistic framework, but the manuscript itself notes that Eq. (2) is a 'ranking-oriented surrogate.' The likelihood P(C|X_j) is a latent cosine similarity and the prior P(X_j) is a hand-designed distinctiveness/sparsity term; the final α_j in Eq. (5) is a rank-fusion score, not a posterior probability. This is not necessarily an empirical flaw, but it is an overclaim in the abstract and introduction. Please either replace the Bayesian language with a direct statement of the proposed scoring heuristics or provide a genuine derivation from the stated generative model.
  5. [Table 1] Table 1 is internally inconsistent. The header promises columns for ArtBench-2 and ArtBench-5, each with Validation and Generation and both 10 and 100 settings, but the non-parametric rows and the FrED row contain only four numeric entries each. In particular, FrED has no ArtBench-5 values, despite the text claiming results on both splits. Please correct the table, clarify what the '10' and '100' columns mean, and report the missing numbers.
minor comments (4)
  1. [§2] Typo: 'KGa' should be 'KGs'.
  2. [§1] The paper-structure paragraph ends with '6.' — remove the stray numeral.
  3. [§5] The conclusion says FrED is 'competitive with several parametric estimators,' which is fair, but the abstract and introduction should make clear that the headline gain over simple baselines comes primarily from latent-space sharpening, not from the KG component, given the ablation evidence.
  4. [§4.2] The environmental study is clearly labeled as analog retrieval rather than causal attribution. This is a reasonable framing, but the abstract's phrase 'training data attribution' could mislead readers; consider using 'domain-grounded analog retrieval' in the abstract as well.

Circularity Check

1 steps flagged · score 4.0 of 10

One self-confirming metric in the flood case study; the core ArtBench LDS evaluation is independently benchmarked.

  1. self definitional [§4.2.1 'Direct Geospatial Matching'; §4.2.2 'Geographic Precision'; Tables 8-9]
    "Unlike the art study’s contrastive inference, here we utilise Direct Geospatial Matching. Since forecasts provide precise coordinates and timestamps, we query the exact environmental state from the KG. ... Geographic Precision: We quantify the percentage of top-K historical analogs that share the target forecast’s country."

    The environmental KG stores the same country/region information used as the evaluation target: Table 8 lists the edge 'Disaster hasCountry Region' and Table 9 gives Region nodes 'name, latitude, longitude'. The Node2Vec domain embeddings are trained on this graph topology, so R_dom already encodes country/regional identity. Re-ranking by R_dom and then measuring 'geographic precision' as same-country overlap therefore largely verifies that the graph encoded its own country labels, not that KG grounding independently adds physical consistency. The reported 65.7% vs 50.9% improvement is substantially a function of the input labels, making the metric self-confirming rather than an external validation.

full rationale

The central ArtBench claim is not circular. The LDS is computed against retraining-based counterfactual ground truth under the D-TRAK protocol, and the attribution score in Eqs. (5)-(6) is evaluated against that external target; no fitted value is renamed as a prediction. The paper’s own ablation shows that most of FrED’s gain over CLIP cosine comes from non-linear scaling of the visual encoder (visual-only p(c|xj)^10 p(xj) = 28.5 vs proposed 29.1), which weakens the KG-grounding narrative and raises a credit-assignment concern, but that is a correctness/attribution issue, not a circular derivation. The contrastive projection anchor subset is underspecified and its OOD generalization is unreported, a reproducibility risk but not a cycle. Self-citations [20,23,24] appear as related work, a dataset source, and a limitations note, not as load-bearing justification. The only concrete circularity is the flood-case geographic precision metric, which uses the KG’s own country/region labels as both re-ranking input and evaluation target. Because that is a secondary feasibility study and the main ArtBench result is independently benchmarked, the overall score is moderate rather than high.

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

FrED introduces no new physical or theoretical entity. Its free parameters are the sharpening exponent, the boost constants W and k, the contrastive temperature, and an underspecified scarcity prior. The domain KGs are constructed from external sources and are not hat-pulled entities, but they are used as a grounding signal without a formal proof that graph topology corresponds to influence.

free parameters (5)
  • Sharpening exponent p = 10
    Chosen from the ablation in Table 7 to maximize LDS on ArtBench; drives most of the improvement over baselines.
  • Boost max scaling W = not reported
    Controls the maximum weight of the domain boost in Eq. (6); value never given in the paper.
  • Boost decay constant k = not reported
    Controls how quickly the domain boost decays with rank in Eq. (6); value never given.
  • Contrastive temperature tau = learned, initialized at 0.07
    Learnable temperature in the CLIP-style InfoNCE alignment loss in Appendix B.2.
  • Informational-scarcity prior p(X_j) = undefined
    Described verbally as a density penalty but never given a concrete functional form; its behavior is implicitly tuned through the final LDS results.
assumptions (5)
  • standard math Bayes' theorem as used in Eq. (3) to expand the attribution weight alpha_j.
    The algebra of Bayes' rule is standard; the issue is the meaning of the substituted terms, not the rule itself.
  • domain assumption The generative model factorizes into an encoder E and decoder D, and target C is a decoded latent state (Section 3.1).
    Used to justify tracing attribution through latent proximity; reasonable for diffusion and forecasting models but not established for all black-box models.
  • ad hoc to paper The target can be approximated as C ≈ Σ_j α_j X_j (Eq. 2).
    The authors explicitly call this a 'ranking-oriented surrogate' rather than a literal generative model; it is the load-bearing heuristic of the whole framework.
  • ad hoc to paper Latent proximity in a pre-trained encoder is a valid proxy for P(C|X_j), and KG topology/rarity is a valid proxy for structure-aware likelihood and prior.
    No theoretical argument connects these similarity and sparsity terms to counterfactual training-data influence; this is the core inductive leap of the paper.
  • domain assumption The precomputed D-TRAK models, images, and protocol are directly comparable and correctly reproduced.
    Table 1 compares against parametric baselines under the claimed identical D-TRAK setup, but the paper does not independently verify all baseline assets.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FrED: External Data Influence Estimation via Domain Knowledge Graph Grounding." pith.science (2026). https://pith.science/paper/YLOS74UV

@misc{pith2026260721615,
  author       = {Pith},
  title        = {Pith review of: FrED: External Data Influence Estimation via Domain Knowledge Graph Grounding},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YLOS74UV}},
  note         = {Machine review of arXiv:2607.21615}
}
read the original abstract

The rapid deployment of generative AI has amplified the critical need for Training Data Attribution to ensure transparency and accountability. However, current parametric approaches require computationally prohibitive access to model weights, while similarity-based methods ignore deep structural context. We propose a novel probabilistic framework that operates entirely in a black-box setting. Our method fuses continuous feature similarities with discrete, domain-specific Knowledge Graphs (KGs). This approach ensures the attribution is grounded in structural reality, explicitly rewarding highly specific historical samples while preventing generic background data from dominating the results. We evaluate our framework across two distinct domains where linking outputs to data and domain context is inherently complex: abstract artistic image synthesis and high-dimensional physical weather forecasting. Extensive benchmarking demonstrates the robust efficacy of our approach. In the artistic domain, it achieves a strong Linear Datamodeling Score that exceeds standard black-box similarity baselines, while closing much of the gap to gradient-based estimators. We additionally present a cross-domain feasibility case study in environmental forecasting, where we use domain KGs to retrieve physically consistent historical analogs for regional flood forecasts, improving geographic localisation over a latent-only baseline. Operating entirely without internal model access, our approach provides an efficient, interpretable mechanism for post-hoc influence analysis and domain-grounded retrieval.

Figures

Figures reproduced from arXiv: 2607.21615 by the authors.

Figure 1
Figure 1. FrED’s overview showing the fusion between latent and domain spaces. A hierarchical mechanism applies the discrete domain rank as a contextual boost to the continuous latent score. 4 Experiments We evaluate FrED across two domains with distinct attribution challenges and evaluation paradigms. In Art Generation (Sec. 4.1), we benchmark against parametric methods on ArtBench to trace stylistic lineages. In Environment… view at source ↗
Figure 2
Figure 2. Artistic exemplar: (a) visual data and (b) its extracted KG, demonstrating the mapping of [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Environmental exemplar: (a) meteorological data and (b) its extracted KG, demonstrating [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: The artwork Ruokokoski Maalaa (a) presented alongside an image of its creator, Sallinen [PITH_FULL_IMAGE:figures/full_fig_p013_4.png]
Figure 5
Figure 5. Figure 5: Ontological schema of the Artistic KG (GD), defining the structural relationships used to map ArtBench to its domain context [PITH_FULL_IMAGE:figures/full_fig_p014_5.png]
Figure 6
Figure 6. Figure 6: Graph visualisation of the painting Ruokokoski Maalaa (a) and the artist Sallinen Tyko (b). [PITH_FULL_IMAGE:figures/full_fig_p016_6.png]
Figure 7
Figure 7. Figure 7: The contrastive multimodal alignment procedure. A batch of raw continuous data (images) [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]
Figure 8
Figure 8. Figure 8: Algorithmic implementation of the contrastive alignment loop. The procedure leverages par [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]
Figure 9
Figure 9. Figure 9: The ontological schema of the Environmental KG ( [PITH_FULL_IMAGE:figures/full_fig_p023_9.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

46 extracted references · 2 linked inside Pith

  1. [23]

    Training data attribu- tion for image generation using ontology-aligned knowledge graphs.arXiv preprint arXiv:2512.02713, 2025

    Theodoros Aivalis, Iraklis A Klampanos, Antonis Troumpoukis, and Joemon M Jose. Training data attribu- tion for image generation using ontology-aligned knowledge graphs.arXiv preprint arXiv:2512.02713, 2025

    arXiv 2025

  2. [1]

    Understanding black-box predictions via influence functions, 2020

    Pang Wei Koh and Percy Liang. Understanding black-box predictions via influence functions, 2020

    2020

  3. [2]

    Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act), 2024

    European Parliament and Council. Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act), 2024. Official Journal of the European Union, L 1689, 12.7.2024, p. 1–60

    2024

  4. [3]

    UK National AI Strategy, 2021

    UK Government. UK National AI Strategy, 2021. Published by Department for Science, Innovation and Technology and Office for Artificial Intelligence and Department for Digital, Culture, Media and Sport and Department for Business, Energy and Industrial Strategy

    2021

  5. [4]

    Recommendation on the Ethics of Artificial Intelligence, 2021

    UNESCO. Recommendation on the Ethics of Artificial Intelligence, 2021. Document code: SHS/BIO/REC- AIETHICS/2021, 21 pages

    2021

  6. [5]

    State of the global climate 2025

    World Meteorological Organization. State of the global climate 2025. Technical Report WMO-No. 1391, World Meteorological Organization (WMO), Geneva, 2026

    2025

  7. [6]

    The copernicus programme and its climate change service

    Jean-Noël Thépaut, Dick Dee, Richard Engelen, and Bernard Pinty. The copernicus programme and its climate change service. InIGARSS 2018 - 2018 IEEE International Geoscience and Remote Sensing Symposium, pages 1591–1593, 2018

    2018

  8. [7]

    Learning skillful medium-range global weather forecasting.Science, 382(6677):1416–1421, 2023

    Remi Lam, Alvaro Sanchez-Gonzalez, Matthew Willson, Peter Wirnsberger, Meire Fortunato, Ferran Alet, Suman Ravuri, Timo Ewalds, Zach Eaton-Rosen, Weihua Hu, et al. Learning skillful medium-range global weather forecasting.Science, 382(6677):1416–1421, 2023

    2023

Show all 46 references
  1. [8]

    Accurate medium-range global weather forecasting with 3d neural networks.Nature, 619(7970):533–538, 2023

    Kaifeng Bi, Lingxi Xie, Hengheng Zhang, Xin Chen, Xiaotao Gu, and Qi Tian. Accurate medium-range global weather forecasting with 3d neural networks.Nature, 619(7970):533–538, 2023

    2023

  2. [9]

    Training data influence analysis and estimation: a survey.Machine Learning, 113(5):2351–2403, March 2024

    Zayd Hammoudeh and Daniel Lowd. Training data influence analysis and estimation: a survey.Machine Learning, 113(5):2351–2403, March 2024

    2024

  3. [10]

    Leave-one-out unfairness, 2021

    Emily Black and Matt Fredrikson. Leave-one-out unfairness, 2021

    2021

  4. [11]

    The shapley taylor interaction index

    Mukund Sundararajan, Kedar Dhamdhere, and Ashish Agarwal. The shapley taylor interaction index. In Hal Daumé III and Aarti Singh, editors,Proceedings of the 37th International Conference on Machine Learning, volume 119 ofProceedings of Machine Learning Research, pages 9259–926...

    2020

  5. [12]

    why should i trust you?

    Marco Tulio Ribeiro, Sameer Singh, and Carlos Guestrin. "why should i trust you?": Explaining the predictions of any classifier, 2016

    2016

  6. [13]

    Gomez, Lukasz Kaiser, and Illia Polosukhin

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. Attention is all you need, 2023

    2023

  7. [14]

    Transformer interpretability beyond attention visualization

    Hila Chefer, Shir Gur, and Lior Wolf. Transformer interpretability beyond attention visualization. pages 782–791, 06 2021. 10

    2021

  8. [15]

    Data attribution for text-to-image models by unlearning synthesized images

    Sheng-Yu Wang, Aaron Hertzmann, Alexei A Efros, Jun-Yan Zhu, and Richard Zhang. Data attribution for text-to-image models by unlearning synthesized images. InThe Thirty-eighth Annual Conference on Neural Information Processing Systems, 2024

    2024

  9. [16]

    Trak: attributing model behavior at scale

    Sung Min Park, Kristian Georgiev, Andrew Ilyas, Guillaume Leclerc, and Aleksander M ˛ adry. Trak: attributing model behavior at scale. InProceedings of the 40th International Conference on Machine Learning, ICML’23. JMLR.org, 2023

    2023

  10. [17]

    Intriguing properties of data attribution on diffusion models

    Xiaosen Zheng, Tianyu Pang, Chao Du, Jing Jiang, and Min Lin. Intriguing properties of data attribution on diffusion models. InThe Twelfth International Conference on Learning Representations, 2024

    2024

  11. [18]

    Diffusion attribution score: Evaluating training data influence in diffusion models

    Jinxu Lin, Linwei Tao, Minjing Dong, and Chang Xu. Diffusion attribution score: Evaluating training data influence in diffusion models

  12. [19]

    Laurens van der Maaten and Geoffrey E. Hinton. Visualizing data using t-sne.Journal of Machine Learning Research, 9:2579–2605, 2008

    2008

  13. [20]

    Enhancing inter- pretability in generative ai through search-based data influence analysis

    Theodoros Aivalis, Iraklis A Klampanos, Antonis Troumpoukis, and Joemon M Jose. Enhancing inter- pretability in generative ai through search-based data influence analysis. InICLR 2025 Workshop on Navigating and Addressing Data Problems for Foundation Models

    2025

  14. [21]

    Fast data attribution for text-to-image models

    Sheng-Yu Wang, Aaron Hertzmann, Alexei A Efros, Richard Zhang, and Jun-Yan Zhu. Fast data attribution for text-to-image models. InThe Thirty-ninth Annual Conference on Neural Information Processing Systems, 2025

    2025

  15. [22]

    Efros, Jun-Yan Zhu, and Richard Zhang

    Sheng-Yu Wang, Alexei A. Efros, Jun-Yan Zhu, and Richard Zhang. Evaluating data attribution for text-to-image models. InICCV, 2023

    2023

  16. [24]

    Llm-driven knowledge graph construction from earth observation data for extreme events

    Theodoros Aivalis, Iraklis A Klampanos, and Antonis Troumpoukis. Llm-driven knowledge graph construction from earth observation data for extreme events. 2025

    2025

  17. [25]

    MIT press Cambridge, MA, USA, 2017

    Yoshua Bengio, Ian Goodfellow, Aaron Courville, et al.Deep learning, volume 1. MIT press Cambridge, MA, USA, 2017. Sections 14.1 (Autoencoders) and 15.1 (Representation Learning)

    2017

  18. [26]

    Springer, 2006

    Christopher M Bishop and Nasser M Nasrabadi.Pattern recognition and machine learning, volume 4. Springer, 2006. Section 1.2.1 (Sum and Product Rules)

    2006

  19. [27]

    Cormack, Charles L A Clarke, and Stefan Buettcher

    Gordon V . Cormack, Charles L A Clarke, and Stefan Buettcher. Reciprocal rank fusion outperforms condorcet and individual rank learning methods. InProceedings of the 32nd International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR ’09, page 7...

    2009

  20. [28]

    The artbench dataset: Benchmarking generative models with artworks.arXiv preprint arXiv:2206.11404, 2022

    Peiyuan Liao, Xiuyu Li, Xihui Liu, and Kurt Keutzer. The artbench dataset: Benchmarking generative models with artworks.arXiv preprint arXiv:2206.11404, 2022

    2022 arXiv

  21. [29]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. Learning transferable visual models from natural language supervision. In Marina Meila and ...

    2021

  22. [30]

    Reproducible scaling laws for contrastive language-image learning

    Mehdi Cherti, Romain Beaumont, Ross Wightman, Mitchell Wortsman, Gabriel Ilharco, Cade Gordon, Christoph Schuhmann, Ludwig Schmidt, and Jenia Jitsev. Reproducible scaling laws for contrastive language-image learning. InProceedings of the IEEE/CVF conference on computer vision ...

    2023

  23. [31]

    node2vec: Scalable feature learning for networks

    Aditya Grover and Jure Leskovec. node2vec: Scalable feature learning for networks. InProceedings of the 22nd ACM SIGKDD international conference on Knowledge discovery and data mining, pages 855–864, 2016

    2016

  24. [32]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bjorn Ommer. High-resolution image synthesis with latent diffusion models. pages 10674–10685, 06 2022

    2022

  25. [33]

    LoRA: Low-rank adaptation of large language models

    Edward J Hu, yelong shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. LoRA: Low-rank adaptation of large language models. InInternational Conference on Learning Representations, 2022. 11

    2022

  26. [34]

    Trak: attributing model behavior at scale

    Sung Min Park, Kristian Georgiev, Andrew Ilyas, Guillaume Leclerc, and Aleksander M ˛ adry. Trak: attributing model behavior at scale. InProceedings of the 40th International Conference on Machine Learning, pages 27074–27113, 2023

    2023

  27. [35]

    Input similarity from the neural network perspective

    Guillaume Charpiat, Nicolas Girard, Loris Felardos, and Yuliya Tarabalka. Input similarity from the neural network perspective. InNeurIPS 2019-33th Annual Conference on Neural Information Processing Systems, 2019

    2019

  28. [36]

    Estimating training data influence by tracing gradient descent

    Garima Pruthi, Frederick Liu, Satyen Kale, and Mukund Sundararajan. Estimating training data influence by tracing gradient descent. In H. Larochelle, M. Ranzato, R. Hadsell, M.F. Balcan, and H. Lin, editors, Advances in Neural Information Processing Systems, volume 33, pages 1...

    2020

  29. [37]

    Identifying a training-set attack’s target using renormalized influence estimation

    Zayd Hammoudeh and Daniel Lowd. Identifying a training-set attack’s target using renormalized influence estimation. InProceedings of the 2022 ACM SIGSAC Conference on Computer and Communications Security, pages 1367–1381, 2022

    2022

  30. [38]

    The journey, not the destination: How data guides diffusion models

    Kristian Georgiev, Joshua Vendrow, Hadi Salman, Sung Min Park, and Aleksander Madry. The journey, not the destination: How data guides diffusion models. InICML Workshop on Challenges in Deployable Generative AI, 2023

    2023

  31. [39]

    Relatif: Identifying explanatory training samples via relative influence

    Elnaz Barshan, Marc-Etienne Brunet, and Gintare Karolina Dziugaite. Relatif: Identifying explanatory training samples via relative influence. InInternational Conference on Artificial Intelligence and Statistics, pages 1899–1909. PMLR, 2020

    1909

  32. [40]

    inaturalist research-grade observations, 2026

    iNaturalist. inaturalist research-grade observations, 2026. Accessed: May 2026

    2026

  33. [41]

    Esa worldcover 10 m 2021 v200

    Daniele Zanaga, Ruben Van De Kerchove, Dirk Daems, Wanda De Keersmaecker, Carsten Brockmann, Grit Kirches, Jan Wevers, Oliver Cartus, Maurizio Santoro, Steffen Fritz, et al. Esa worldcover 10 m 2021 v200. 2022

    2021

  34. [42]

    Weatherbench 2: A benchmark for the next generation of data-driven global weather models, 2023

    Stephan Rasp, Stephan Hoyer, Alexander Merose, Ian Langmore, Peter Battaglia, Tyler Russel, Alvaro Sanchez-Gonzalez, Vivian Yang, Rob Carver, Shreya Agrawal, Matthew Chantry, Zied Ben Bouallegue, Peter Dueben, Carla Bromberg, Jared Sisk, Luke Barrington, Aaron Bell, and Fei Sh...

    2023

  35. [43]

    Model-agnostic interpretability of machine learning.arXiv preprint arXiv:1606.05386, 2016

    Marco Tulio Ribeiro, Sameer Singh, and Carlos Guestrin. Model-agnostic interpretability of machine learning.arXiv preprint arXiv:1606.05386, 2016. A Artistic KG: Schema and Construction To establish the topological prior for the artistic domain, we constructed a comprehensive ...

    2016 arXiv

  36. [44]

    Primary Ingestion:We initiated the pipeline using painting identifiers and artist names provided by ArtBench-10

  37. [45]

    We then performed a secondary expansion to ingest granular biographical profiles for the associated artists, including active years and thematic tags

    Metadata Expansion (WikiArt):Using these identifiers as primary keys, we queried the WikiArt API to systematically retrieve all available technical and categorical metadata for each artwork. We then performed a secondary expansion to ingest granular biographical profiles for t...

  38. [46]

    a {class} painting

    Relational Resolution (Wikipedia & DBpedia):To capture all the possible historical context, we utilised the Wikipedia URLs embedded within the WikiArt records as bridge identifiers to aggregate data from both Wikipedia and DBpedia. From Wikipedia, we extracted available biogra...

    2018

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.