Pith. sign in

REVIEW 4 major objections 5 minor 29 references

A Dynamic-Semantics Framework for Grounding Human Referring Expressions in Visual Perceptual Data

T0 review · 4 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read The paper argues that explicit, inspectable pact-state bookkeeping — three sets of referent-object bindings updated by dynamic-semantics rules — is the missing component for machine grounding, and shows that one utterance puts the right…

desk verdict A well-meant, transparent component paper whose central formal claim is undercut by an internal set-operation bug and a missing multi-turn experiment. read the letter →

arxiv 2608.08663 v1 pith:HBIA5OLU submitted 2026-08-09 cs.HC cs.CV

Joseph Bingham This is my paper · ORCID
classification cs.HCcs.CV
keywords lexicalentrainmentconceptualpactsdynamicsemanticsreferringexpressionsmultimodalgroundingvisualtangramreferencegameUniversalQualityIndex
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 argues that the missing ingredient for machines that must interpret human referring expressions is not better perception but explicit bookkeeping of what the listener and speaker have agreed to call each object. It proposes a listener-side layer that maintains three sets of pact bindings — committed ($\Gamma$), under negotiation ($\Xi$), and ruled out ($\Omega$) — and updates them with a dynamic-semantics context-change rule after every utterance. Stacked on a perceptual channel that aligns web-retrieved images to abstract silhouettes, the framework places the correct target in its top-5 hypothesis set 83.56% of the time from a single utterance, or 67.8% after removing retrievals that could be memorized. The defensible claim is deliberately narrow: a transparent, auditable symbolic component for pact tracking, not a full interactive entrainment system.

What carries the argument

The load-bearing object is the triple $\Gamma, \Xi, \Omega$ together with the update rule $C[\varphi] = C \cap \Diamond B$, where $B$ is the set of candidate referent-object bindings surviving perceptual alignment. $\Diamond$ (might) adds candidates to $\Xi$, $\square$ (must) commits a unique binding to $\Gamma$, and $\square\neg$ moves rejected bindings into $\Omega$; the rule makes the common ground inspectable as three explicit sets rather than an opaque embedding. The perceptual layer supplies $B$ by converting each utterance into a search query, retrieving images, aligning them to each tangram with a scale-invariant feature transform homography, and thresholding the Universal Quality Index comparison, with rotation and grayscale-inversion augmentation handling mirror and polarity variation.

What would settle it

Re-run the pipeline with a retrieval corpus built to exclude all images resembling the target tangrams; if top-5 accuracy falls to the 8.33% chance level under that exclusion, the reported grounding signal is an artifact of retrieval memorization rather than cross-modal grounding.

Watch

Extended reading notes

Core claim

The central discovery is that lexical entrainment's pact state can be externalized into three explicit sets — $\Gamma$ for finalized commitments, $\Xi$ for hypotheses still under negotiation, and $\Omega$ for rejected bindings — with a context-change-potential update rule that narrows $\Xi$ by intersection across turns and promotes a binding to $\Gamma$ once a single hypothesis survives. On top of this symbolic layer, a perceptual pipeline estimates the hypothesis set for each utterance by retrieving crowd-sourced images, aligning them to each candidate with a scale-invariant feature transform, and scoring alignment quality with the Universal Quality Index. The paper shows this combination recovers the structure of lexical entrainment turn by turn and grounds a single director utterance with 83.56% top-5 accuracy, dropping to 67.8% under a conservative leakage filter designed to strip out retrieved copies of the target itself.

Load-bearing premise

The evaluation assumes that the web images retrieved for each phrase, after preprocessing, fairly sample what a human matcher would visualize, so that alignment between those images and a tangram indicates real grounding rather than retrieval of the corpus itself.

Editorial extensions

If this is right

  • If the framework is right, vision-language agents can carry an inspectable, auditable record of what has been agreed with a partner, directly addressing the absence of stable pact state documented in current neural systems.
  • The 83.56% top-5 result, and the 67.8% figure after conservative filtering, imply that a genuine cross-modal grounding signal survives even when obvious memorization routes are removed; both are well above the 8.33% chance level.
  • The ablation results isolate the contribution of each component, so future systems know that query preprocessing, SIFT alignment, UQI scoring, and rotation/inversion augmentation each earn their place.
  • The result is explicitly a component, not a closed system: a full entrainment model would need a generation-side partner model and interactive repair, which this paper does not attempt.

Reading between the lines

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

  • A reader could expect the $\Gamma/\Xi/\Omega$ layer to transfer directly to other referent types — actions, plans, abstract concepts — because the update semantics is agnostic about what the similarity function measures; the bottleneck is the perceptual channel, not the bookkeeping.
  • Because the commit rule is deterministic and monotone, the framework would fail to recover from an early false commitment; adding probabilistic commit thresholds is a natural extension that the paper itself flags.
  • The conservative filter is a proxy for leakage; a stronger test would compare against a learned perceptual similarity baseline and use a retrieval corpus constructed to exclude the target stimulus class entirely, which the paper identifies as the most informative next experiment.
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 proposes a symbolic listener-side framework for grounding referring expressions in the Stanford Repeated Reference Game corpus. A dynamic-semantics layer maintains three sets of pact bindings (Γ, Ξ, Ω) updated by context-change rules, and a perceptual layer converts director utterances into web-image queries, aligns retrieved images to tangram candidates via SIFT homographies, and scores them with UQI. The authors report top-5 single-utterance accuracy of 83.56% (unfiltered) and 67.8% under a conservative leakage filter, with top-1 accuracy of 41.66% and 29.2%, respectively, against a human top-1 baseline of roughly 77–80%. They also present ablations of the retrieval pipeline and an explicit internal-validity discussion of retrieval leakage, and they position the contribution as an inspectable, auditable pact-state bookkeeping layer that could be integrated into interactive emergent-communication or RSA-style systems.

Significance. If the claims were fully supported, the paper would supply a useful, transparent component for a problem that recent LVLM studies identify as a gap: stable, inspectable pact-state tracking. The manuscript is unusually honest: it reports the previously missing ablation tables, includes a conservative leakage-filtered analysis, explicitly quantifies the retrieval-leakage threat, and releases code. The main contribution is the symbolic Γ/Ξ/Ω layer together with a reproducible perceptual-alignment pipeline. However, the significance as stated is limited by two gaps: the formal update rules as written do not obviously implement accumulation of pacts, and the empirical evaluation is single-utterance only, so the central 'turn by turn recovery' claim is not demonstrated. With appropriate revisions to the formalism, the experiment, and the claims, the component could be a valuable contribution to multimodal grounding and emergent-communication research.

major comments (4)
  1. [Section 3.4, Eqs. (1)–(3); Section 3.5; Section 2.7] Under the stated interpretation of Γ as 'the set of pacts the MCP has committed to,' the update Γ ← Γ ∩ □(rφ ← o_i) is internally inconsistent with accumulation: intersecting a set of pacts with a singleton deletes every prior pact. The same issue affects Ω in Eq. (3), and the Ξ update in Eq. (2) similarly removes rather than accumulates. Section 3.5's worked example uses union-like language ('Γ gains □(rφ3 ← o_A)'), and Section 2.7 says the layer 'accumulates hypothesized pacts over turns.' If the ∩ notation is intended as a context-of-worlds update, then Γ/Ξ/Ω are not sets of pacts and the paper's 'inspectable pact state' description is misleading. This is the central formal contribution, and the inconsistency must be resolved before the symbolic layer can be assessed.
  2. [Abstract and Section 1 vs. Sections 4, 5.2, 6.3] The abstract and introduction claim that the framework 'recovers the structure of lexical entrainment turn by turn,' but the experiments in Tables 1–3 and Section 4 evaluate only single director utterances; no multi-turn sequence is passed through the Γ/Ξ/Ω bookkeeping. Sections 5.2 and 6.3 explicitly concede that the system is not interactive and that no recovery mechanism exists. The 'turn by turn' claim is therefore asserted rather than demonstrated. Either remove it from the abstract and contribution list, or add an actual multi-turn evaluation on the repeated-reference corpus that tracks commitment, refinement, and pruning across consecutive utterances.
  3. [Sections 4.2, 4.3, and 6.4] All accuracy numbers are point estimates from a single retrieval pass, and the hyperparameters (k, the UQI threshold ε, the augmentation configuration, the similarity metric, and the 0.85 conservative-filter threshold) are selected by performance on the same evaluation corpus. The paper acknowledges variance but still reports 83.56% and 67.8% as headline figures. Without bootstrap confidence intervals, multiple independent retrieval runs, or a held-out hyperparameter split, the empirical claims are not quantitatively robust; the conservative condition, in particular, is defined by a threshold that is itself tuned on the data being scored.
  4. [Table 3 and Section 5.1] The comparison of MCP top-5 accuracy (83.56%) with the human top-1 baseline (77–80%) is an apples-to-oranges comparison: human top-k scores for k > 1 are not available, and top-1 is a stricter success criterion than top-5. The statement in Section 5.1 that the unfiltered top-5 'matches or slightly exceeds the human top-1 baseline' should be either removed or accompanied by the explicit caveat that the two metrics are not directly comparable, and the abstract should not juxtapose the 83.56% top-5 figure with the human top-1 figure without that caveat.
minor comments (5)
  1. [Table 1 caption] The caption contains a typo: 'T able 1' should read 'Table 1.'
  2. [Section 2.4] The sentence 'The common ground is the set of possible worlds consistent with Γ∩Ξ∩Ω' is unclear because Γ, Ξ, and Ω are defined as sets of pacts, not sets of possible worlds; the formal type of each set and the meaning of '∩' between them should be clarified.
  3. [Section 5.3] The 16% and 32% gaps used as proxies for utterances that cannot be grounded are acknowledged as proxies, but the text should state explicitly that these are estimates derived from the accuracy gaps, not from a direct annotation of ungroundable utterances.
  4. [Reference [15]] The citation 'Anonymous: Lvlms and humans ground differently in referential communication' should be replaced with the full author list and venue information if available, so that readers can locate the work.
  5. [Table 3] The note 'Chance on 12 tangrams is 8.33%' applies to top-1 accuracy only; for top-k accuracy with k>1, chance-level performance is k/12 and should be reported separately to help readers interpret the top-3 and top-5 numbers.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper reports empirical measurements with explicit ablations and leakage bounds, and its formal update layer is stipulated rather than fitted to the evaluation outcome.

full rationale

The paper's derivation chain is empirical rather than definitional. The top-k accuracies in Section 4.4 are measurements of a fixed retrieval-and-alignment pipeline, and Section 4.3 reports a conservative filtering condition rather than renaming the unfiltered result as a prediction. The formal layer in Section 3.4 is stipulated as an update semantics, not fitted to the accuracy numbers; its update rules (Gamma intersection, Xi difference, Omega intersection) do not contain the evaluation outcome as an input. The hyperparameter choices (k=7, UQI+SIFT, the appended 'tangram figure' keyword, and the 0.85 filter) are selected by performance on the evaluation corpus and reported as ablations with explicit point-estimate caveats; this is selection on the test data and a validity risk, but it is parameter fitting rather than a prediction forced by construction, so it does not meet the standard for a circularity finding. The paper explicitly disclaims the strong 'reproduces entrainment' interpretation in Sections 5.2 and 6.1, and the internal inconsistency between the intersection update in Eq. (1) and the 'Gamma gains' language in Section 3.5 is a formal bug, not a circular reduction. No load-bearing self-citation or imported uniqueness theorem appears: references [10] and [11] supply notation, not the paper's result, and the empirical baselines in [5] and [9] are external. I therefore find no step in which an output is equivalent by definition to an input.

Assumptions & free parameters 6 free parameters · 4 assumptions · 1 invented entities

The empirical claims depend on the retrieval query construction, the similarity scoring chain, the leakage filter, and the corpus comparison, none of which is derived from the dynamic-semantics layer. Several of these components are tuned on the evaluation corpus, and the Gamma, Xi, Omega bookkeeping is a stipulated model rather than a consequence of prior theory.

free parameters (6)
  • k (number of web images retrieved per query) = 7
    Chosen from a 1-10 scan in Figure 5; accuracy peaks at k=6 or 7, so k=7 is used in the final pipeline. This selection on the evaluation corpus is a fitted hyperparameter.
  • epsilon (decision threshold for similarity) = rank-based, no numeric value given
    The paper states epsilon is the key hyperparameter (Section 2.5) and a rank-based threshold performed 3 to 5 points better than absolute cutoffs in 0.3, 0.4, 0.5, 0.6, 0.7 (Section 4.2). The rank mechanism is chosen by performance, so the effective decision boundary is fit to data.
  • appended query keyword 'tangram figure' = present
    Hand-chosen domain-specific augmentation to every query (Section 3.2). It is essential (raw queries at chance) but also a direct contributor to leakage; this is a manually fitted design choice, not a parameter with a numeric value.
  • image augmentation configuration = rotation + grayscale inversion
    Selected by ablation (Table 1): no augmentation 33.3%, rotation 38.1%, inversion 36.4%, both 41.66%. The final configuration is chosen because it maximizes the reported accuracy on the same corpus.
  • conservative leakage filter threshold = 0.85 on normalized UQI scale
    Manual filter in Section 4.3 to remove likely target-duplicate retrievals; the threshold is chosen by hand and affects the headline conservative numbers (top-5 drops to 67.8%).
  • similarity metric plus alignment combination = UQI + SIFT
    Selected from ten metrics in Table 2 because it maximizes top-1 accuracy. This is a model-selection choice made on the evaluation corpus.
assumptions (4)
  • domain assumption Dynamic semantics with context-change potentials and epistemic modals is an adequate formal model for listener-side pact state.
    The paper stipulates that common ground can be represented as three sets Gamma, Xi, Omega updated by intersection operations (Sections 2.3 and 2.4). This is a modeling choice, not derived from the data or from prior formal results.
  • domain assumption Bing image search results, after POS filtering and appending 'tangram figure', sample the human visual prior that directors use to produce referring expressions.
    The perceptual channel's validity depends on this premise (Sections 2.5 and 3.2). The paper flags leakage but does not establish that non-leaked retrievals are representative of human imagery.
  • domain assumption SIFT homography alignment followed by UQI measures the perceptual similarity between a tangram silhouette and a retrieved web image.
    The scoring chain g(o_i, I_phi) assumes this (Section 3.3). The ablation shows it is the best among ten classical metrics, but does not validate it against human perceptual judgments.
  • domain assumption The Stanford Repeated Reference Game corpus is a valid testbed for evaluating listener-side pact tracking, and the human top-1 baselines from Hawkins et al. transfer to the MCP's top-k setup.
    Section 4.1 and Table 3 assume corpus transcripts and human accuracy figures are a fair benchmark for a machine that sees object sets but not matcher dialog. Human top-k is unavailable, so the comparison is approximate.
invented entities (1)
  • Gamma, Xi, Omega pact-state sets
    purpose: Represent listener-side common ground as three sets of referent-object bindings: finalized pacts, pacts under negotiation, and ruled-out pacts.
    This tripartite formal construct is new to this paper and has no falsifiable handle outside the paper; its utility is judged through the paper's own evaluation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Dynamic-Semantics Framework for Grounding Human Referring Expressions in Visual Perceptual Data." pith.science (2026). https://pith.science/paper/HBIA5OLU

@misc{pith2026260808663,
  author       = {Pith},
  title        = {Pith review of: A Dynamic-Semantics Framework for Grounding Human Referring Expressions in Visual Perceptual Data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HBIA5OLU}},
  note         = {Machine review of arXiv:2608.08663}
}
abstract

Humans converge on shared names for novel, hard-to-describe objects through repeated interaction, a process psycholinguists call lexical entrainment. Leading vision-language models fail at this: recent empirical work documents that they do not shorten references, reuse successful expressions, or maintain stable pact state across turns. We present a framework that addresses the gap by externalizing pact state into three explicit, inspectable sets of referent-object bindings ($\Gamma, \Xi, \Omega$), updated by a dynamic-semantics context-change rule. The symbolic layer sits on top of a lightweight perceptual-alignment pipeline that grounds noisy human referring expressions in crowd-sourced imagery via SIFT homographies and the Universal Quality Index. Evaluated on the Stanford Repeated Reference Game corpus (over 15{,}000 director-matcher utterances on abstract tangram stimuli), the framework places the correct target in its top-5 hypothesis set 83.56% of the time from a single director utterance. Human matcher top-1 accuracy on the same corpus is approximately 77-80%. We also report results on a held-out condition in which obvious tangram-adjacent images are excluded from the retrieved set, which provides a more conservative measurement of the grounding signal. Ablations isolate the contribution of each component: SIFT alignment, UQI, query preprocessing, and image augmentation. The central contribution is the combination: a transparent, auditable symbolic layer that recovers the structure of lexical entrainment turn by turn, paired with a perceptual channel whose behavior can be examined ablation by ablation. We also discuss in detail what the framework does not do. It is not interactive, it does not close the loop with the director, and its retrieval-driven perceptual channel is vulnerable to a class of leakage effects that we quantify and bound rather than wave away.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

29 extracted references · 23 canonical work pages

  1. [1]

    Linguistics and philosophy25(5/6), 701–721 (2002)

    Stalnaker, R.: Common ground. Linguistics and philosophy25(5/6), 701–721 (2002)

    work page 2002

  2. [2]

    Journal of experimental psychology: Learning, memory, and cognition22(6), 1482 (1996)

    Brennan, S.E., Clark, H.H.: Conceptual pacts and lexical choice in conversation. Journal of experimental psychology: Learning, memory, and cognition22(6), 1482 (1996)

    work page 1996

  3. [3]

    Brown, R.: Words and things. (1958)

    work page 1958

  4. [4]

    Trainin, N., Shetreet, E.: We do not speak like this here: The role of perceived foreignness in shaping speaker-specific social and linguistic inferences. Cogn. Sci. 49(7), 70086 (2025)

    work page 2025

  5. [5]

    Cogni- tive Science44(6), 12845 (2020) https://doi.org/10.1111/cogs.12845 https://onlinelibrary.wiley.com/doi/pdf/10.1111/cogs.12845

    Hawkins, R.D., Frank, M.C., Goodman, N.D.: Characterizing the dynamics of learning in repeated reference games. Cogni- tive Science44(6), 12845 (2020) https://doi.org/10.1111/cogs.12845 https://onlinelibrary.wiley.com/doi/pdf/10.1111/cogs.12845

    doi:10.1111/cogs.12845 2020

  6. [6]

    arXiv preprint arXiv:2006.02419 (2020) https://doi.org/10.48550/ ARXIV.2006.02419

    Lazaridou, A., Baroni, M.: Emergent multi-agent communication in the deep learning era. arXiv preprint arXiv:2006.02419 (2020) https://doi.org/10.48550/ ARXIV.2006.02419

    arXiv doi:10.48550/arxiv.2006.02419 2020

  7. [7]

    Trends in Cognitive Sciences7(7), 308–312 (2003)

    Steels, L.: Evolving grounded communication for robots. Trends in Cognitive Sciences7(7), 308–312 (2003)

    work page 2003

  8. [8]

    In: Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence (IJCAI-25), pp

    Kouwenhoven, T., Peeperkorn, M., Kleijn, R., Verhoef, T.: Shaping shared languages: Human and large language models’ inductive biases in emergent com- munication. In: Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence (IJCAI-25), pp. 10298–10306 (2025). https://doi.org/10. 24963/ijcai.2025/1144 .https://arxiv.org/ab...

    arXiv 2025

Show all 29 references
  1. [9]

    In: Proceedings of the 39th Conference on Neural Information Processing Systems (NeurIPS) Workshop: Dynamics of Behavior and Mind (2025).https://openreview.net/forum?id=0xWIIC0Ibh

    Zhao, H., Conwell, C.: Comparing human and machine communication patterns through a tangram game. In: Proceedings of the 39th Conference on Neural Information Processing Systems (NeurIPS) Workshop: Dynamics of Behavior and Mind (2025).https://openreview.net/forum?id=0xWIIC0Ibh

    2025

  2. [10]

    Mind128(511), 795–835 (2019) https://doi.org/10.1093/mind/fzy076

    Goldstein, S.: Generalized update semantics. Mind128(511), 795–835 (2019) https://doi.org/10.1093/mind/fzy076

    2019 doi

  3. [11]

    PhD thesis, Rutgers, The State University of New Jersey–New Brunswick, New Brunswick, NJ, USA (2017)

    Goldstein, S.: Informative dynamic semantics. PhD thesis, Rutgers, The State University of New Jersey–New Brunswick, New Brunswick, NJ, USA (2017). https://doi.org/10.7282/T3PR804Q

    2017 doi

  4. [12]

    pypi (Feb 10, 2022)

    Singh, G.P.: bing-image-downloader 1.1.2. pypi (Feb 10, 2022). https://pypi.org/ project/bing-image-downloader/

    2022

  5. [13]

    In: Meila, M., Zhang, T

    Zhu, H., Neubig, G., Bisk, Y.: Few-shot language coordination by mod- eling theory of mind. In: Meila, M., Zhang, T. (eds.) Proceedings of the 38th International Conference on Machine Learning. Proceedings of Machine Learning Research, vol. 139, pp. 12901–12911. PMLR, ??? (202...

    2021

  6. [14]

    arXiv preprint arXiv:2509.03805 (2025) https://doi

    Imai, S., ˙Inan, M., Sicilia, A., Alikhani, M.: Measuring how (not just whether) vlms build common ground. arXiv preprint arXiv:2509.03805 (2025) https://doi. org/10.48550/ARXIV.2509.03805

    2025 doi arXiv

  7. [15]

    arXiv preprint arXiv:2601.19792 (2026)

    Anonymous: Lvlms and humans ground differently in referential communication. arXiv preprint arXiv:2601.19792 (2026). Full author list per arXiv metadata

    2026 arXiv

  8. [16]

    In: Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems

    Maeda, K., McCarthy, W.P., Tsai, C.-Y., Mu, J., Wang, H., Hawkins, R.D., Fan, J.E., Abtahi, P.: Gesturing toward abstraction: Multimodal convention formation in collaborative physical tasks. In: Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems. CHI ...

    2026

  9. [17]

    arXiv preprint arXiv:2407.17489 (2024) https://doi.org/10.48550/ARXIV.2407.17489

    Riedl, C., Savage, S., Zvelebilova, J.: Cognitive spillover in human-ai teams. arXiv preprint arXiv:2407.17489 (2024) https://doi.org/10.48550/ARXIV.2407.17489

    2024 doi

  10. [18]

    Evolution of Communica- tion1(1), 1–34 (1997)

    Steels, L.: The synthetic modeling of language origins. Evolution of Communica- tion1(1), 1–34 (1997)

    1997

  11. [19]

    In: Experiments in Cultural Language Evolution vol

    Steels, L., Loetzsch, M.: The grounded naming game. In: Experiments in Cultural Language Evolution vol. 3, pp. 41–59. John Benjamins, ??? (2012)

    2012

  12. [20]

    PLoS ONE8(3), 58960 (2013)

    Beuls, K., Steels, L.: Agent-based models of strategies for the emergence and evolution of grammatical agreement. PLoS ONE8(3), 58960 (2013)

    2013

  13. [21]

    IEEE Access11, 142317–142340 (2023)

    Brandizzi, N.: Toward more human-like ai communication: A review of emergent 22 communication research. IEEE Access11, 142317–142340 (2023)

    2023

  14. [22]

    Transactions on Machine Learning Research (2024)

    Boldt, B., Mortensen, D.R.: A review of the applications of deep learning-based emergent communication. Transactions on Machine Learning Research (2024)

    2024

  15. [23]

    Language Development Research 5(1), 116–143 (2025)

    Galke, L.P.A., Raviv, L.: Emergent communication and learning pressures in lan- guage models: A language evolution perspective. Language Development Research 5(1), 116–143 (2025)

    2025

  16. [24]

    In: Raviv, L., Boeckx, C

    Rita, M., Michel, P., Chaabouni, R., Pietquin, O., Dupoux, E., Strub, F.: Lan- guage evolution with deep learning. In: Raviv, L., Boeckx, C. (eds.) The Oxford Handbook of Approaches to Language Evolution. Oxford Handbooks. Oxford University Press, ??? (2025)

    2025

  17. [25]

    To appear (2017)

    Honnibal, M., Montani, I.: spaCy 2: Natural language understanding with Bloom embeddings, convolutional neural networks and incremental parsing. To appear (2017)

    2017

  18. [26]

    arXiv (2021)

    Zhao, Y., Huang, X., Zhang, Z.: Deep Lucas-Kanade Homography for Multimodal Image Alignment. arXiv (2021). https://doi.org/10.48550/ARXIV.2104.11693 . https://arxiv.org/abs/2104.11693

    2021 doi arXiv

  19. [27]

    Scholarpedia7(5), 10491 (2012) https://doi.org/10.4249/scholarpedia.10491

    Lindeberg, T.: Scale Invariant Feature Transform. Scholarpedia7(5), 10491 (2012) https://doi.org/10.4249/scholarpedia.10491 . revision #153939

    2012 doi

  20. [28]

    IEEE Signal Processing Letters9(3), 81–84 (2002) https://doi.org/10.1109/97.995823

    Wang, Z., Bovik, A.C.: A universal image quality index. IEEE Signal Processing Letters9(3), 81–84 (2002) https://doi.org/10.1109/97.995823

    2002 doi

  21. [29]

    arXiv (2019)

    Hawkins, R.D., Frank, M.C., Goodman, N.D.: Characterizing the dynamics of learning in repeated reference games. arXiv (2019). https://doi.org/10.48550/ ARXIV.1912.07199 . https://arxiv.org/abs/1912.07199 23

    2019 doi arXiv

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.