Pith. sign in

REVIEW 4 major objections 4 minor 42 references

From Direction to Magnitude: How Multimodal Instruction-Tuning Reorganizes the Geometric Encoding of Identity-Specifying Prompts in Transformer Hidden States

T0 review · 4 major / 4 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read Multimodal instruction-tuning moves identity-specifying prompts from a direction-coded to a magnitude-coded geometric fingerprint in transformer hidden states.

desk verdict A careful, honest empirical paper whose headline direction-to-magnitude claim outruns its evidence — the norm inversion is never content-controlled, and the borderline Euclidean p fails FDR. read the letter →

arxiv 2607.09842 v2 pith:2EYHUTS5 submitted 2026-07-10 cs.LG cs.CL

Jorge A. Castillo , Marco Torres Yévenes , Juan Carlos Lanas This is my paper · ORCID
classification cs.LGcs.CL
keywords hidden-stategeometryOllivier-RiccicurvatureWassersteindistanceidentity-specifyingpromptsinstructiontuningdirectionvsmagnitudeencodingtransformerinterpretabilitytrajectoryanalysis
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

This paper asks whether identity-specifying system prompts leave a distinguishable geometric trace in transformer hidden-state trajectories once prompt length and generated content are controlled. Across a four-model panel spanning no training, multimodal RLHF, RL distillation, and SFT, the authors find a regime-dependent reorganization: in the untrained base model the identity fingerprint lives in the direction of hidden-state vectors, while in the multimodal instruction-tuned model it moves into the magnitude. The identity-vanilla separation collapses under angular (direction-only) comparison but survives under Euclidean comparison, and the norm of the first generated state is lowest for the identity prompt even though it is the longest prompt. The effect is specific to the multimodal instruction-tuning regime and is concentrated in the first generated token, with a teacher-forced content control attributing about 30 percent of the free-running cosine signal to the prompt itself. If correct, the finding implies that identity information in post-trained models may be carried by an overlooked channel—vector magnitude—and that the encoding substrate is not fixed across training regimes.

What carries the argument

The argument is carried by comparing the same point cloud of hidden states under two base metrics. Trajectories of 256 hidden states are turned into k-NN graphs; for each edge, Ollivier-Ricci curvature measures how much the random-walk neighborhoods of the two endpoints overlap under optimal transport, and the per-condition pooled distribution of edge curvatures is compared by the 1-Wasserstein distance. Running the same protocol with Euclidean distance versus angular (L2-normalized) distance isolates whether a geometric difference lives in vector magnitude or in direction. A teacher-forced control, in which different prompts are forced to generate the same target tokens, separates prompt-dr

What would settle it

Run the norm measurement with teacher-forced shared targets: condition an identity prompt and a generic prompt to generate the same neutral target sequences. If the identity prompt no longer yields the lowest norm of the first generated state—or if a length- and format-matched non-identity prompt does—the magnitude-coding claim is an artifact of template surface features rather than identity content.

Watch

Extended reading notes

Core claim

The central empirical claim is a qualitative reorganization of identity encoding across the instruction-tuning boundary. In the base-weight model, the axis-vs-vanilla separation under edge-wise Ollivier-Ricci curvature on k-NN trajectory graphs survives and strengthens under angular normalization (separation 0.034, permutation p=0.002), so identity is encoded in the direction of hidden states. In the multimodal instruction-tuned model, the same separation collapses under angular k-NN (p=0.439) but persists under Euclidean k-NN (p=0.042 at B=5000), and the mean norm of the first generated state is lowest for the identity prompt (138.9 vs 211.5 generic and 195.3 vanilla)—an inversion of the le

Load-bearing premise

The load-bearing premise is that the lowered norm of the first generated hidden state under the identity prompt comes from its identity content and not from surface features such as prompt length, formatting, or the position of the last token, since the norm result was not checked under the prompt-to-content control the paper applies to its other signals.

Editorial extensions

If this is right

  • If the reorganization is real, identity-related information in instruction-tuned models is not stored along the directional axes that steering and probing methods usually search for; norm-based statistics would be needed to read it.
  • The length confound is directly refuted: the longest prompt produces the smallest first-state norm in the tuned model, so the magnitude signal tracks semantic content rather than prompt length.
  • The prompt's geometric effect is concentrated in the transition from the final prompt token to the first generated token and decays within roughly ten tokens; trajectory-level clustering mostly reflects generated content.
  • The direction-to-magnitude switch is specific to multimodal RLHF among the four regimes tested; RL distillation retains length-driven separation and SFT shows none, so the change cannot be attributed to instruction-tuning in general.
  • The methodological combination—Wasserstein distance on edge-wise Ollivier-Ricci curvature distributions—provides a graph-level way to compare full trajectory geometries, which the paper argues is independently useful.

Reading between the lines

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

  • If norm-coding is genuine, hidden-state magnitude could be an unguarded channel: interventions that normalize or steer directions would leave identity content intact, and detection methods should add norm-based features.
  • The inversion is consistent with instruction-tuning learning to compress a long identity template into a single low-norm 'mode' at the first token; a layer-resolved trace should show where the norm drop is introduced, a direct next experiment.
  • The roughly 30 percent prompt-driven residual at the first token suggests identity conditioning acts like a switch that changes the initial state's norm; one testable consequence is that deliberately manipulating that norm could modulate identity-related behavior without changing token-level content.
  • Because only one model pair instantiates the multimodal regime, the regime-specificity claim would be stress-tested by repeating the angular-vs-Euclidean diagnostic on other base/tuned pairs; if the inversion fails to appear there, the finding narrows to this specific model family.
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 / 4 minor

Summary. The paper asks whether identity-specifying system prompts induce statistically distinguishable geometric fingerprints in transformer hidden-state trajectories, and how these depend on the post-training regime. Using four 7–8B open-weight models spanning four regimes (base, multimodal RLHF, RL distillation, SFT), it compares three prompt conditions via several geometric statistics, with the primary metric being the 1-Wasserstein distance between edge-wise distributions of Ollivier–Ricci curvature on k-NN trajectory graphs. The central claim is a direction-to-magnitude reorganization across the instruction-tuning boundary: in base Gemma-4-E4B the identity signal survives angular normalization (directional encoding), while in Gemma-4-E4B-it it collapses under angular normalization and is instead detected in Euclidean space and in the norm of the first generated state. The paper includes trajectory-level permutation tests, sensitivity sweeps, an FDR audit, and a teacher-forced content control.

Significance. If the direction-to-magnitude reorganization is real, it is a noteworthy qualitative finding about how instruction tuning changes the geometric substrate of prompt encoding, and the proposed W1-on-edge-Ollivier-Ricci statistic is a potentially useful methodological tool. The paper is unusually careful in its inferential discipline: permutation tests are trajectory-level, borderline p-values are re-run at B=5000, BH-FDR and Holm corrections are reported, k-NN and PCA sensitivities are swept, anisotropy is addressed via ABT projection, and teacher-forced controls isolate content-driven effects. The authors also disclose their limitations explicitly. My assessment is that the statistical reporting is honest and mostly sound, but the magnitude-coding half of the central claim currently rests on a norm inversion that has not been subjected to the paper's own content/template controls, while the only direct geometric evidence for it is a single FDR-failing p-value.

major comments (4)
  1. [§4.1.2, Table 4] Table 4 is the principal evidence for magnitude coding of identity in Gemma-4-E4B-it: the axis prompt yields mean ||v1||=138.9 vs 211.5 for generic and 195.3 for vanilla. However, this statistic is never included in the teacher-forced protocol of §4.1.4/Table 5, and no template/format control is reported. The axis and generic prompts differ not only in identity content but in length (2129 vs 957 tokens), wording, formatting, and final-token position; the paper's 'refutation of length as confound' only rules out a monotonic length effect, not a template or special-token artifact. If the inversion disappears under teacher-forced shared targets or under a length-matched/format-matched control, the magnitude half of the central claim collapses. Please report ||v1|| in the teacher-forced setting and with at least one format-controlled generic variant.
  2. [Table 3 and Table 9] The only non-norm geometric evidence for magnitude coding in the instruction-tuned model is the Euclidean axis-vs-vanilla W1 with p=0.042 at B=5000. As the paper correctly states, this fails BH-FDR at q=0.05 (rank 8, threshold 0.040). The paper nevertheless treats 'survives under Euclidean k-NN' as a central corroboration in §4.1.2 and §6. Given the FDR failure, this comparison is at best supportive; the central claim therefore depends even more heavily on the norm inversion of Table 4, whose controls are missing. Please either strengthen this comparison or explicitly restate the central claim as resting on the norm channel only.
  3. [§4.1.4, Table 5] The teacher-forced control shows that the free-running Euclidean curvature signal W1(rho_axis,rho_vanilla) attenuates by ~94% (0.0288 to 0.0016) and becomes non-significant. Thus most of the IT-model curvature separation is content-driven, not prompt-driven. The ~30% 'prompt-driven component proper' is estimated only for the cosine statistic Delta C01 and is not shown to apply to the norm channel. Since the norm inversion is not teacher-forced, it could reflect differences in the content generated by the prompts rather than a prompt-driven geometric fingerprint. Please either report the teacher-forced ||v1|| or limit the magnitude-coding claim to free-running trajectories.
  4. [§7, Future work] The paper's abstract and §1.4 state that the direction-to-magnitude reorganization is 'specific to the multimodal instruction-tuning regime,' but each regime is represented by a single model. As the authors acknowledge in §7, multiple models per regime are needed to support a regime-specificity claim. Without replication within a regime, the observed difference could be due to architecture, data mix, or model family rather than the post-training regime. Please temper the regime-specificity language in the abstract and introduction, or add additional checkpoints per regime.
minor comments (4)
  1. [§1.4, H2 and §6] The phrase 'normative substrate' appears to be a typo; the text elsewhere uses 'magnitude' or 'norm'. Please replace 'normative' with 'norm-based' or 'magnitude'.
  2. [Table 2 vs Table 3] The p-value for W1(rho_axis,rho_vanilla) in Gemma-4-E4B-it is reported as 0.042 in Table 2 and as 0.047 (refined to 0.042 at B=5000) in Table 3, while Table 5 lists 0.047 for what appears to be the same free-running comparison. Please harmonize these values and clarify which one is the headline p-value.
  3. [§2.2] The generic prompt is called 'length-matched' but its length (957 tokens) is not matched to the axis prompt (2129 tokens). Consider describing it as 'length-controlled' or 'similarly long' to avoid confusion.
  4. [Appendix C] The pooled-edge W1 uses ~60,000 curvature values per condition while the permutation test shuffles only 100 trajectories. The per-trajectory variant in Appendix C is a useful check, but the main text could explicitly note that the pooled statistic does not treat edges as independent samples for inference.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity; the central empirical claim is measured under an external permutation protocol, not derived from fitted inputs.

full rationale

The paper's central claims are empirical and self-contained against the reported measurements. No constants are fitted to produce the reported Wasserstein distances, permutation p-values, silhouette scores, or norm means; the trajectory-level permutation nulls are external to the observed values, and the sensitivity sweeps (k, d_PCA, ABT, B=5000) are reported rather than selected to force significance. The direction-to-magnitude reorganization is inferred from a pre-specified methodological contrast -- Euclidean vs angular k-NN graph construction (Section 3.1) -- rather than from a parameter fitted to the outcome. The norm inversion (Table 4) is a direct measurement, and the paper itself flags the borderline FDR status of the Euclidean p=0.042 (Table 9, Section 5). The only self-referential element is the use of the authors' own VEX/MIA template as the axis prompt; this functions as an experimental stimulus, not as a load-bearing citation, and the geometric conclusions do not reduce to any property asserted by that template. The disclosure that H2 was 'enriched ... by an explicit prediction concerning the geometric substrate' is a transparency caveat about hypothesis formation, not a circular reduction: the predictions are stated and then tested with the same external permutation protocol, and no fitted coefficient or equation is reused as evidence. Accordingly no specific circular step meets the evidentiary bar of quoting a constructional equivalence.

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

The paper fits no free parameters to data; the quantitative claims are permutations and descriptive statistics. The ledger therefore records the hand-chosen analysis hyperparameters and the domain assumptions that carry the interpretation. The most load-bearing choice is accepting the 957-token generic prompt as a length/content control for a 2129-token axis prompt, which the paper acknowledges is asymmetric.

free parameters (5)
  • k-NN neighborhood size k = 5
    Chosen by heuristic k≈log(256); sensitivity over {5,10,15,20} reported in Table 6.
  • Ollivier idleness α = 1/2
    Follows prior work (Sandhu et al. 2015; Topping et al. 2022); not swept.
  • Trajectory length N = 256
    Fixed for tractability; the paper conjectures stability across N in [128,512] but does not verify.
  • PCA dimension d_PCA = 50
    Default for silhouette; sensitivity over {30,50,100,200} reported.
  • Permutation count B = 1000 (5000 for borderline)
    B=5000 used for the borderline IT Euclidean p-value; higher B would refine further.
assumptions (5)
  • domain assumption Final-layer hidden state at the latest generated-token position is a sufficient probe of identity encoding.
    All metrics use only the last pre-lm_head layer (Section 3.1, Table 7 justification); layer-resolved analysis is deferred.
  • domain assumption k-NN geometry on N=256 samples in intrinsic dimension ~10 approximates local manifold structure.
    Relies on convergence results of Van der Hoorn et al. and Trillos-Weber; the paper cites these, but the approximation is not verified here.
  • ad hoc to paper The 100 internally curated ontology prompts are representative of identity-specifying queries.
    The prompt set is internal and no external replication benchmark is provided (Appendix A, Limitations).
  • standard math Greedy decoding at temperature 0 yields deterministic trajectories; token-indexed sequence is a point cloud with temporal ordering.
    Definition 2.1 and Section 1.3 framing.
  • ad hoc to paper The generic prompt (957 tokens) is an adequate length/content control for the axis prompt (2129 tokens).
    The 2.2x length disparity is acknowledged but the control is still used to attribute differences to identity content (Section 2.2, H1).

how reviews work

0 comments
Cite this review

Pith. "Pith review of From Direction to Magnitude: How Multimodal Instruction-Tuning Reorganizes the Geometric Encoding of Identity-Specifying Prompts in Transformer Hidden States." pith.science (2026). https://pith.science/paper/2EYHUTS5

@misc{pith2026260709842,
  author       = {Pith},
  title        = {Pith review of: From Direction to Magnitude: How Multimodal Instruction-Tuning Reorganizes the Geometric Encoding of Identity-Specifying Prompts in Transformer Hidden States},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2EYHUTS5}},
  note         = {Machine review of arXiv:2607.09842}
}
read the original abstract

We investigate whether identity-specifying system prompts produce statistically distinguishable geometric fingerprints in the hidden-state trajectories of four open-weight transformer language models spanning four post-training regimes: no training (Gemma-4-E4B base), multimodal RLHF (Gemma-4-E4B-it), RL distillation (DeepSeek-R1-Distill-Qwen-7B), and SFT (Qwen2.5-7B-Instruct). Three prompt conditions (an identity-specifying axis prompt, a length-matched generic-assistant prompt, and a 26-token vanilla baseline) are compared via five geometric metrics, principally the 1-Wasserstein distance between edge-wise distributions of Ollivier-Ricci curvature on k-NN trajectory graphs. Claims rest on trajectory-level permutation tests with multiple geometric controls (teacher-forced content controls, temporal-chain vs k-NN topology, ABT-projected k-NN, angular vs Euclidean graph construction, B=5000 permutations on borderline statistics). The central finding is a qualitative reorganization of identity encoding across the instruction-tuning boundary: in the base model the fingerprint is direction-coded (separation 0.034, p=0.002 under angular k-NN); in the multimodal instruction-tuned model it migrates into the magnitude (angular separation collapses to p=0.439 while Euclidean survives at p=0.042, and the mean norm of the first generated state inverts its length-ordering, being lowest for the identity prompt). This direction-to-magnitude reorganization is specific to the multimodal instruction-tuning regime, absent under RL distillation and SFT. A teacher-forced control attributes ~30% of the free-running cosine signal to prompt-driven effects. We position W_1 on edge-wise Ollivier-Ricci distributions on k-NN trajectory graphs as a methodological contribution of independent interest.

Figures

Figures reproduced from arXiv: 2607.09842 by the authors.

Figure 1
Figure 1. Trajectory centroids projected onto the first three principal components of the pooled [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 5 linked inside Pith

  1. [1]

    Alvarez-Melis, D., and Jaakkola, T. S. (2018). Gromov--Wasserstein alignment of word embedding spaces. EMNLP

    2018

  2. [2]

    M., and Perona, I

    Arbelaitz, O., Gurrutxaga, I., Muguerza, J., Pérez, J. M., and Perona, I. (2013). An extensive comparative study of cluster validity indices. Pattern Recognition, 46(1):243--256

    2013

  3. [3]

    Belinkov, Y., and Glass, J. (2019). Analysis methods in neural language processing: a survey. TACL, 7:49--72

    2019

  4. [4]

    Belrose, N., Furman, Z., Smith, L., Halawi, D., McKinney, I., Ostrovsky, Y., Biderman, S., and Steinhardt, J. (2023). Eliciting latent predictions from transformers with the tuned lens. arXiv:2303.08112

    arXiv 2023

  5. [5]

    Chen, R., Arditi, A., Sleight, H., Evans, O., and Lindsey, J. (2025). Persona vectors: monitoring and controlling character traits in language models. arXiv:2507.21509

    arXiv 2025

  6. [6]

    Cohen-Steiner, D., Edelsbrunner, H., and Harer, J. (2007). Stability of persistence diagrams. Discrete & Computational Geometry, 37:103--120

    2007

  7. [7]

    Cohen-Steiner, D., Edelsbrunner, H., Harer, J., and Mileyko, Y. (2010). Lipschitz functions have L_p -stable persistence. Foundations of Computational Mathematics, 10:127--139

    2010

  8. [8]

    M., and Scherf, N

    Torbati, N., Gaebler, M., Hofmann, S. M., and Scherf, N. (2025). Geometry matters: insights from Ollivier-Ricci curvature and Ricci flow into representational alignment. arXiv:2501.00919

    arXiv 2025

Show all 42 references
  1. [9]

    Elhage, N., Nanda, N., Olsson, C., Henighan, T., Joseph, N., Mann, B., et al. (2021). A mathematical framework for transformer circuits. Transformer Circuits Thread

    2021

  2. [10]

    Ethayarajh, K. (2019). How contextual are contextualized word representations? Comparing the geometry of BERT, ELMo, and GPT-2 embeddings. EMNLP-IJCNLP

    2019

  3. [11]

    Facco, E., d'Errico, M., Rodriguez, A., and Laio, A. (2017). Estimating the intrinsic dimension of datasets by a minimal neighborhood information. Scientific Reports, 7:12140

    2017

  4. [12]

    T., Tannenbaum, A., and Lenglet, C

    Farooq, H., Chen, Y., Georgiou, T. T., Tannenbaum, A., and Lenglet, C. (2019). Network curvature as a hallmark of brain structural connectivity. Nature Communications, 10:4937

    2019

  5. [13]

    Forman, R. (2003). Bochner's method for cell complexes and combinatorial Ricci curvature. Discrete & Computational Geometry, 29(3):323--374

    2003

  6. [14]

    Fournier, N., and Guillin, A. (2015). On the rate of convergence in Wasserstein distance of the empirical measure. Probability Theory and Related Fields, 162:707--738

    2015

  7. [15]

    Hewitt, J., and Manning, C. D. (2019). A structural probe for finding syntax in word representations. NAACL-HLT

    2019

  8. [16]

    Kantorovich, L. V. (1942). On the translocation of masses. Doklady Akademii Nauk SSSR, 37:199--201

    1942

  9. [17]

    Rautenstrauch, P., and Ohler, U. (2025). Shortcomings of silhouette in single-cell integration benchmarking. Nature Biotechnology. DOI: 10.1038/s41587-025-02743-4

    2025 doi

  10. [18]

    Levina, E., and Bickel, P. J. (2004). Maximum likelihood estimation of intrinsic dimension. NeurIPS

    2004

  11. [19]

    Lu, C., Gallagher, J., Michala, J., Fish, K., and Lindsey, J. (2026). The Assistant Axis: situating and stabilizing the default persona of language models. arXiv:2601.10387

    2026

  12. [20]

    Mu, J., Bhat, S., and Viswanath, P. (2018). All-but-the-top: simple and effective postprocessing for word representations. ICLR

    2018

  13. [21]

    Nanda, N., Chan, L., Lieberum, T., Smith, J., and Steinhardt, J. (2023). Progress measures for grokking via mechanistic interpretability. ICLR

    2023

  14. [22]

    Nguyen, K., et al. (2023). Revisiting over-smoothing and over-squashing using Ollivier-Ricci curvature. ICML

    2023

  15. [23]

    Ollivier, Y. (2009). Ricci curvature of Markov chains on metric spaces. Journal of Functional Analysis, 256(3):810--864

    2009

  16. [24]

    Peyré, G., and Cuturi, M. (2019). Computational optimal transport. Foundations and Trends in Machine Learning, 11(5--6)

    2019

  17. [25]

    Razzhigaev, A., Mikhalchuk, M., Goncharova, E., Oseledets, I., Dimitrov, D., and Kuznetsov, A. (2024). The shape of learning: anisotropy and intrinsic dimensions in transformer-based models. Findings of EACL

    2024

  18. [26]

    Rousseeuw, P. J. (1987). Silhouettes: a graphical aid to the interpretation and validation of cluster analysis. Journal of Computational and Applied Mathematics, 20:53--65

    1987

  19. [27]

    P., Gu, J., Liu, S., Saucan, E., and Jost, J

    Samal, A., Sreejith, R. P., Gu, J., Liu, S., Saucan, E., and Jost, J. (2018). Comparative analysis of two discretizations of Ricci curvature for complex networks. Scientific Reports, 8:8650

    2018

  20. [28]

    Sandhu, R., Georgiou, T., Reznik, E., Zhu, L., Kolesov, I., Senbabaoglu, Y., and Tannenbaum, A. (2015). Graph curvature for differentiating cancer networks. Scientific Reports, 5:12323

    2015

  21. [29]

    Santambrogio, F. (2015). Optimal Transport for Applied Mathematicians. Birkhäuser

    2015

  22. [30]

    P., Mohanraj, K., Jost, J., Saucan, E., and Samal, A

    Sreejith, R. P., Mohanraj, K., Jost, J., Saucan, E., and Samal, A. (2016). Forman curvature for complex networks. Journal of Statistical Mechanics, 2016:063206

    2016

  23. [31]

    Tenney, I., Das, D., and Pavlick, E. (2019). BERT rediscovers the classical NLP pipeline. ACL

    2019

  24. [32]

    Timkey, W., and van Schijndel, M. (2021). All bark and no bite: rogue dimensions in transformer language models obscure representational quality. EMNLP

    2021

  25. [33]

    P., Dong, X., and Bronstein, M

    Topping, J., Di Giovanni, F., Chamberlain, B. P., Dong, X., and Bronstein, M. M. (2022). Understanding over-squashing and bottlenecks on graphs via curvature. ICLR

    2022

  26. [34]

    García Trillos, N., and Weber, M. (2023). Continuum limits of Ollivier's Ricci curvature on data clouds: pointwise consistency and global lower bounds. arXiv:2307.02378

    2023 arXiv

  27. [35]

    J., Mini, U., and MacDiarmid, M

    Turner, A., Thiergart, L., Leech, G., Udell, D., Vazquez, J. J., Mini, U., and MacDiarmid, M. (2023). Activation addition: steering language models without optimization. arXiv:2308.10248

    2023 arXiv

  28. [36]

    Valeriani, L., Doimo, D., Cuturello, F., Laio, A., Ansuini, A., and Cazzaniga, A. (2023). The geometry of hidden representations of large transformer models. NeurIPS

    2023

  29. [37]

    Van der Hoorn, P., Cunningham, W., Lippner, G., Trugenberger, C., and Krioukov, D. (2023). Ollivier-Ricci curvature convergence in random geometric graphs. Discrete & Computational Geometry

    2023

  30. [38]

    N., Kaiser, L., and Polosukhin, I

    Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., and Polosukhin, I. (2017). Attention is all you need. NeurIPS

    2017

  31. [39]

    Villani, C. (2009). Optimal Transport: Old and New. Springer

    2009

  32. [40]

    Wang, Z. (2025). The geometry of persona: disentangling personality from reasoning in large language models. arXiv:2512.07092

    2025

  33. [41]

    Weed, J., and Bach, F. (2019). Sharp asymptotic and finite-sample rates of convergence of empirical measures in Wasserstein distance. Bernoulli, 25(4A):2620--2648

    2019

  34. [42]

    Zou, A., et al. (2023). Representation engineering: a top-down approach to AI transparency. arXiv:2310.01405

    2023 arXiv

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.