What's in a prompt? Language models encode literary style in prompt embeddings

Christopher Earls; Haley Moller; Nicolas Boull\'e; Rapha\"el Sarfati; Toni J. B. Liu

REVIEW 4 major objections 5 minor 32 references

What's in a prompt? Language models encode literary style in prompt embeddings

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Short literary excerpts carry a stylistic signature readable from a single deep LLM embedding.

desk verdict A careful probe study showing style signals in LLM prompt embeddings, but the style-vs-topic confound is not controlled; the specific N/L surfaces and cross-lingual transfer are the real novelties. read the letter →

arxiv 2505.17071 v1 pith:GBTCRHS6 submitted 2025-05-19 cs.CL

Raphaël Sarfati , Haley Moller , Toni J. B. Liu , Nicolas Boullé , Christopher Earls This is my paper · ORCID

classification cs.CL

keywords promptembeddingsliterarystyleauthorshipattributiontransformerinternalrepresentationslast-tokenembeddinggeometryprincipalcomponentscross-lingualtransfer

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 sets out to show that when a large language model reads a short passage of prose, the internal vector of the passage's final token has absorbed information about who wrote it, not just what it means. Using chunks of 8 to 128 tokens from 19th- and early 20th-century novels, the authors probe the rightmost embedding after each transformer layer and find that a linear classifier can separate two authors with over 90% accuracy, while a multiclass probe distinguishes among thirteen books with roughly 75% accuracy. The same-author books are much harder to tell apart than different-author books, which the paper takes as evidence that the encoded signal is authorial style rather than topic. If this is right, authorship becomes readable from a single embedding of about a hundred tokens, without any task-specific fine-tuning.

What carries the argument

The central object is the rightmost-token embedding of a chunk of $N$ tokens after transformer layer $L$, written $\vec{x}_N^{(L)}$; because the attention mask is causal, this single vector has accumulated information from every preceding token and serves as the prompt's summary at that depth. The authors collect one such vector per chunk for each novel, forming ensembles, and test whether the ensembles are separable using linear SVM probes for binary comparisons and MLP probes for multiclass comparisons. Two observations carry the argument: separability increases with both context length $N$ and layer depth $L$, and classifier confusion is systematically higher between books by the same author than between books by different authors. PCA truncation, block shuffling, and French-to-English transfer are used as controls to locate the signal in a low-dimensional subspace and to test whether syntax or surface language is responsible.

What would settle it

Train the same probe on two books by the same author after replacing all proper nouns and content nouns with placeholders while keeping syntax and function words intact; if accuracy drops to chance, the separability was carried by factual content, not style. A second check is to split the chunk ensembles so that entire contiguous blocks of a novel go to either training or test rather than being interleaved; if accuracy collapses under that split, the result depends on autocorrelation between neighboring chunks rather than on a stable style signature.

Watch

Extended reading notes

Core claim

The paper's central claim is that deep transformer embeddings encode a stylistic signature of the prompt. After enough transformer layers (about $L \ge 10$) and with enough context (about $N \ge 64$), the rightmost token's activation in Llama-3.2-1B contains linearly readable information about the literary source of a short excerpt. This signature occupies a low-dimensional subspace of the largest principal components (around 16 directions), persists when token order is scrambled at block sizes 1, 4, and 32, and transfers across languages: a probe trained on distinguishing French novelists keeps most of its accuracy on the English translations of the same books. The higher confusion between different novels by the same author than between novels by different authors is the paper's core evidence that what is encoded is style rather than factual content.

Load-bearing premise

The central claim depends on the separability signal being style rather than factual content or shared register; the paper itself allows that names, subjects, and era vocabulary could reveal the source, and because chunks are consecutive windows from one novel, nearby chunks resemble one another in ways that can inflate test accuracy.

Editorial extensions

If this is right

Authorship attribution can work from a single deep embedding of a roughly 100-token excerpt, without fine-tuning a model for the task.
The stylistic signal is carried mainly by lexical content rather than word order, because shuffling tokens in blocks of 1, 4, or 32 leaves probe accuracy high.
The relevant representation is low-dimensional, concentrated in about 16 principal-component directions of the embedding ensemble.
The style signature survives translation between French and English, suggesting it is not tied to the surface language of the text.
The information is built up gradually by the attention mechanism: shallow layers and short contexts do not show separability, while deeper layers and longer contexts do.

Reading between the lines

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

A direct test the paper does not run: compare these deep-embedding probes with a function-word-frequency baseline on the same chunks; if the baseline matches the probe accuracy, the transformer may simply be re-encoding count statistics rather than a new kind of style representation.
The cross-language result hints that style occupies a language-neutral region of the embedding space; one could search for a direction whose subtraction removes authorship information while preserving meaning.
Because different shuffling scales are perfectly separable from one another, syntax is still represented somewhere in the embeddings; ablating the style subspace could show whether this authorship signal is causally used in next-token generation or is a by-product.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Short version: this is a solid empirical probing paper, not a breakthrough. The headline claim—that deep prompt embeddings carry a linearly readable signature of the source oeuvre—is plausible and mostly supported by their measurements, but the "style" interpretation is underdetermined. The paper's own shuffling result (97% accuracy at B=1) shows bag-of-words lexical content is sufficient, so topic and named-entity leakage remain live confounds.

What's actually new: the accuracy-vs-layer-depth-vs-context-length surfaces across four 1B-class base models; the PCA finding that about 16 top principal components carry the separable signal; the same-author confusion pattern; and the cross-lingual transfer experiment. These are concrete, reproducible-ish measurements (no code shipped, but the method is straightforward). The paper also honestly admits in Sec 3.1 that factual information could explain the provenance signal. That's good scholarly hygiene.

Where it's soft: the central confound is not controlled. The 70/30 chunk-level split over consecutive windows from a single novel risks autocorrelation leakage (adjacent chunks share names, settings, discourse context). A held-out-novel evaluation or a function-word-only baseline would have settled a lot. The shuffling experiment is informative but actually strengthens the topic-leakage worry: at B=1, syntax is gone, yet accuracy stays at 97%, so the classifier is mainly using content-word frequencies. Same-author confusion is suggestive, but same-period/same-register authors also share topic distributions (e.g., Melville's sea novels vs. Twain's travelogues share domain vocabulary). No error bars anywhere; with a 70/30 split and random seeds, that's a fixable gap, not a fatal one.

The cross-lingual transfer is the most convincing piece: a probe trained on French Flaubert vs Hugo transfers to the English translations. That suggests something structural rather than surface names, though translation preserves topic as well as style, so it doesn't fully break the tie.

Bottom line: the paper deserves a serious referee, and the gaps are addressable with a moderate revision. For a reading group, it's a useful case study in why binary classifier accuracy alone can't establish a "style" claim. I'd recommend conditional accept with a request for topic-controlled baselines, held-out-novel evaluation, and variance reporting.

Referee Report

4 major / 5 minor

Summary. The paper investigates whether the final-token embeddings produced by transformer language models encode information about the literary source of a prompt. Using chunks of 8–128 tokens drawn consecutively from 19th- and early-20th-century novels, the authors train linear SVM and MLP probes on rightmost-token activations of Llama-3.2-1B and three other base models. They report high binary and multiclass classification accuracy, increased classifier confusion between books by the same author, a low-dimensional PCA subspace that preserves separability, preserved separability under token shuffling, and cross-lingual transfer between French novels and their English translations. The paper concludes that LLM embeddings encode a stylistic signature of the source oeuvre. The central interpretive claim is that the separable signal is style rather than topic or other factual content; the manuscript itself acknowledges this alternative reading in Section 3.1.

Significance. If the stylistic interpretation were established, the paper would be a valuable contribution to LLM interpretability and to authorship attribution, since it would show that intangible prompt-level properties are linearly readable in deep embeddings of short excerpts. The empirical work has real strengths: the use of multiple open models in Appendix D, the explicit attempt to separate syntax from lexical content through shuffling in Section 3.3, and the exploratory cross-lingual transfer experiment in Section 3.4. The paper also promises public code, which is welcome. However, the load-bearing inference from 'classifier separability of novels' to 'embeddings encode style' is not yet supported, because topic, register, and local discourse context are not controlled in the experiments that are offered as evidence. The result as stated is therefore an interesting phenomenon whose interpretation remains underdetermined.

major comments (4)

[Section 3.1, Fig. 3] The same-author confusion pattern is the principal evidence for the stylistic interpretation, but it does not control for topic, register, or period vocabulary. Same-author pairs such as Melville's Moby Dick and Pierre, Twain's Life on the Mississippi and Roughing It, and the three Austen novels share narrative settings, era-specific social vocabulary, and broad thematic domains, so higher confusion between same-author books is exactly what a content-based classifier would produce. The paper needs a control that removes or ablates content words, or an evaluation on held-out novels or held-out authors, before the increased same-author confusion can be attributed to style.
[Section 2 and Appendix B.3] The 70/30 train/test split is applied at the level of individual chunks, and chunks are consecutive non-overlapping windows of N tokens from a single tokenized novel. Neighboring chunks therefore share proper nouns, sentence fragments, discourse context, and local topic, so the test set is autocorrelated with the training set. This leakage can inflate classification accuracy substantially. The authors should either split at the novel or author level, or interleave/downsample chunks to break autocorrelation, and should report the effect of such a split on the accuracies in Figs. 2–3 and Tables 1–2.
[Section 3.3, Table 1] The shuffling experiment shows that at B=1, where word order and syntax are destroyed, the linear probe still achieves 97% accuracy on the GE versus VW pair. Because shuffling preserves the multiset of open-class content words, this result demonstrates that bag-of-words lexical content alone is sufficient for separability; it does not distinguish style from topic. The statement in Section 5 that 'the main signal might be about lexical content rather than syntax' is too weak, given that lexical content is itself confounded with topic and factual references. A function-word-only or content-word-ablated analysis is needed to make the stylistic claim.
[Section 3.4, Table 2] The cross-lingual transfer experiment does not resolve the style-versus-topic confound because translation preserves both stylistic choices and topical content. The observation that a French-trained probe transfers to English translations is consistent with the embeddings encoding topic, named entities, or register, as well as with the embeddings encoding style. The paper should state this limitation explicitly and, if possible, test transfer on a pair of languages or registers where topical overlap is controlled.

minor comments (5)

[Conclusion] The conclusion says 'short (102 tokens) literary excerpts' but the methods and figures use N up to 128; please align the token count or correct the typo.
[Fig. 5 caption] The caption says 'Wolf (VW1, VW2)' but the author name is Woolf; please correct the spelling.
[Table 2] The two rows of accuracies are unlabeled; please indicate which row corresponds to the reference French-trained probe and which to the English-transferred probe, and report the standard errors or confidence intervals.
[Appendix B.3] The classifier description does not specify the number of random seeds, the validation procedure for hyperparameters, or whether the 70/30 split was stratified; these details are needed for reproducibility given the small number of novels.
[Fig. 4B] The label 'GEVW' is unclear; please indicate whether this refers to the combined GE/VW ensemble or to each ensemble separately, and state the embedding layer used for the intrinsic-dimension estimate.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's claims are empirical probe measurements on held-out chunks and out-of-sample cross-lingual transfer, not derivations that reduce to their inputs.

full rationale

The paper's load-bearing claims are empirical: linear and MLP probes are trained on one part of the embedding ensembles and evaluated on held-out chunks; the cross-lingual probe is trained on French pairs and applied to English translations; shuffling ablations are performed and reported as accuracies. None of these steps fits a parameter to the target conclusion and then renames the fit a prediction. The 70/30 chunk-level split means neighboring chunks may be autocorrelated, which is a legitimate validity concern about inflated accuracy, but it is not circularity: the test chunks are still disjoint from training chunks and the reported accuracies are measurements, not consequences of a fitted quantity. The paper itself states the main alternative explanation, 'It could be that they contain enough factual information (names, subjects, etc.) to reveal their provenance' (Section 3.1), and attempts to address it with same-author confusion and shuffling experiments. Whether that evidence isolates style from topic is an interpretative and experimental-design question, not a circularity in the derivation. The only self-citation, Sarfati et al. (2025) in Section 3.2, is used as background motivation for low-dimensional structure; the paper independently recomputes PCA projections and probe accuracy on its own data, so the cited result is not load-bearing. No equation, definition, or fitted parameter is shown to be equivalent by construction to the paper's conclusions. Accordingly, no specific circular step can be exhibited, and the appropriate finding is no significant circularity.

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

No new physical or mathematical entities are introduced. The central claim rests on the validity of causal pooling in the last embedding, the independence of chunks, and the interpretive equation of separability with style. The judged parameters are standard probe architecture choices plus the split ratio; the PCA-to-64 and MLP-32 choices are the main hand-set values.

free parameters (4)

PCA target dimension for SVM probes = 64
Chosen for the binary classifiers before linear SVM; no ablation is shown, and the accuracies in Fig 2 could depend on this choice.
MLP penultimate layer dimension = 32
Arbitrary for the multiclass probe; also used as the UMAP input dimension in Fig 5.
Train/validation split ratio = 70/30
Standard split, but applied to consecutive chunks from the same novel, so it does not guard against autocorrelation leakage.
SVM regularization hyperparameter = not reported
Not stated in the paper; results may depend on it.

assumptions (4)

domain assumption Causal masked attention makes the rightmost token's embedding a function of all preceding tokens in the prompt (Section 2).
Foundation of the method; standard for decoder-only transformers and stated in Section 2.
domain assumption Consecutive N-token chunks from the same novel can be treated as independent samples for train/validation splits.
The 70/30 split is applied to chunks without blocking, but adjacent chunks share local vocabulary, topics, and characters; this assumption is likely violated and affects measured accuracies.
domain assumption Classifier separability, after accounting for same-author confusion, reflects style rather than topic or register.
Interpretive leap in Sections 3.1 and 5; not directly tested.
domain assumption The TwoNN method estimates intrinsic dimension reliably for these embedding ensembles.
Used for Fig 4B, from Valeriani et al. 2023; no validation is provided on this data.

how reviews work

0 comments

Cite this review

Pith. "Pith review of What's in a prompt? Language models encode literary style in prompt embeddings." pith.science (2026). https://pith.science/paper/GBTCRHS6

@misc{pith2026250517071,
  author       = {Pith},
  title        = {Pith review of: What's in a prompt? Language models encode literary style in prompt embeddings},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GBTCRHS6}},
  note         = {Machine review of arXiv:2505.17071}
}

read the original abstract

Large language models use high-dimensional latent spaces to encode and process textual information. Much work has investigated how the conceptual content of words translates into geometrical relationships between their vector representations. Fewer studies analyze how the cumulative information of an entire prompt becomes condensed into individual embeddings under the action of transformer layers. We use literary pieces to show that information about intangible, rather than factual, aspects of the prompt are contained in deep representations. We observe that short excerpts (10 - 100 tokens) from different novels separate in the latent space independently from what next-token prediction they converge towards. Ensembles from books from the same authors are much more entangled than across authors, suggesting that embeddings encode stylistic features. This geometry of style may have applications for authorship attribution and literary analysis, but most importantly reveals the sophistication of information processing and compression accomplished by language models.

Figures

Figures reproduced from arXiv: 2505.17071 by the authors.

**Figure 1.** After semantic embedding of the prompt, vectors represent a single word. As the prompt passes through transformer layers, the attention mechanism funnels more and more information about preceding tokens into the last embedding – turning it into a ‘chimera’ vector, encoding bits of information from all others. This raises the question: What’s in a prompt? In other words, what kind of information contained in the seq… view at source ↗

**Figure 2.** (A) Ensembles of short excerpts (N = 64 tokens) from GE and VW separate in the latent space as embeddings travel through successive transformer layers. (B) Linear classifier accuracy (%) to distinguish GE vs VW ensembles as a function of prompt’s number of tokens N and number of transformer layers crossed L. 3 Results 3.1 Embeddings encode authorship Does a short passage (10-100 words) from a novel contain enough i… view at source ↗

**Figure 4.** Dimensionality of stylistic features. (A) [PITH_FULL_IMAGE:figures/full_fig_p003_4.png] view at source ↗

Figures from the paper (2 more)

**Figure 5.** Figure 5: Map of style: low-dimensional visualization [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]

**Figure 6.** Figure 6: Alternative map of style emphasizing relative [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 9 canonical work pages

[1]

online" 'onlinestring :=

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRING...
[2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...
[3]

Loubna Ben Allal, Anton Lozhkov, Elie Bakouch, Gabriel Martín Blázquez, Guilherme Penedo, Lewis Tunstall, Andrés Marafioti, Hynek Kydlíček, Agustín Piqueres Lajarín, Vaibhav Srivastav, Joshua Lochner, Caleb Fahlgren, Xuan-Son Nguyen, Clémentine Fourrier, Ben Burtenshaw, Hugo Larcher, Haojun Zhao, Cyril Zakka, Mathieu Morlon, and 3 others. 2025. https://ar...

arXiv 2025
[4]

Milad Alshomary, Narutatsu Ri, Marianna Apidianaki, Ajay Patel, Smaranda Muresan, and Kathleen McKeown. 2024. https://arxiv.org/abs/2409.07072 Latent space interpretation for stylistic analysis and explainable authorship attribution . Preprint, arXiv:2409.07072

work page Pith review arXiv 2024
[5]

Trenton Bricken, Adly Templeton, Joshua Batson, Brian Chen, Adam Jermyn, Tom Conerly, Nick Turner, Cem Anil, Carson Denison, Amanda Askell, Robert Lasenby, Yifan Wu, Shauna Kravec, Nicholas Schiefer, Tim Maxwell, Nicholas Joseph, Zac Hatfield-Dodds, Alex Tamkin, Karina Nguyen, and 6 others. 2023. https://transformer-circuits.pub/2023/monosemantic-features...

work page 2023
[6]

Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, and 1 others. 2020. Language models are few-shot learners. In Advances in neural information processing systems, volume 33, pages 1877--1901

work page 2020
[7]

Chang, Zhuowen Tu, and Benjamin K

Tyler A. Chang, Zhuowen Tu, and Benjamin K. Bergen. 2022. https://arxiv.org/abs/2205.10964 The geometry of multilingual language model representations . Preprint, arXiv:2205.10964

arXiv 2022
[8]

Yamshchikov

Svetlana Gorovaia, Gleb Schmidt, and Ivan P. Yamshchikov. 2024. https://doi.org/10.18653/v1/2024.nlp4dh-1.39 S ui generis: Large language models for authorship attribution and verification in L atin . In Proceedings of the 4th International Conference on Natural Language Processing for Digital Humanities, pages 398--412, Miami, USA. Association for Comput...

work page doi:10.18653/v1/2024.nlp4dh-1.39 2024

Show all 32 references

[9]

Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, and 1 others. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783

2024 arXiv
[10]

Wes Gurnee and Max Tegmark. 2024. https://arxiv.org/abs/2310.02207 Language models represent space and time . Preprint, arXiv:2310.02207

2024 arXiv
[11]

Rebecca M. M. Hicke and David Mimno. 2023. https://arxiv.org/abs/2310.18454 T5 meets tybalt: Author attribution in early modern english drama using large language models . Preprint, arXiv:2310.18454

2023 arXiv
[12]

Baixiang Huang, Canyu Chen, and Kai Shu. 2025. https://arxiv.org/abs/2408.08946 Authorship attribution in the era of llms: Problems, methodologies, and challenges . Preprint, arXiv:2408.08946

2025 arXiv
[13]

Artem Kirsanov, Chi-Ning Chou, Kyunghyun Cho, and SueYeon Chung. 2025. https://arxiv.org/abs/2502.08009 The geometry of prompting: Unveiling distinct mechanisms of task adaptation in language models . Preprint, arXiv:2502.08009

2025 arXiv
[14]

Michaud, David D

Yuxiao Li, Eric J. Michaud, David D. Baek, Joshua Engels, Xiaoqing Sun, and Max Tegmark. 2025. https://doi.org/10.3390/e27040344 The geometry of concepts: Sparse autoencoder feature structure . Entropy, 27(4)

2025 doi
[15]

Qing Lyu, Marianna Apidianaki, and Chris Callison-burch. 2023. https://doi.org/10.18653/v1/2023.starsem-1.32 Representation of lexical stylistic features in language models' embedding space . In Proceedings of the 12th Joint Conference on Lexical and Computational Semantics (*...

2023 doi
[16]

Samuel Marks and Max Tegmark. 2024. https://arxiv.org/abs/2310.06824 The geometry of truth: Emergent linear structure in large language model representations of true/false datasets . Preprint, arXiv:2310.06824

2024 arXiv
[17]

MetaAI. 2024. Llama 3.2 model card. https://github.com/meta-llama/llama-models/blob/main/models/llama3_2/MODEL_CARD.md. Accessed: 2024-09-25

2024
[18]

Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013 a . https://arxiv.org/abs/1301.3781 Efficient estimation of word representations in vector space . Preprint, arXiv:1301.3781

2013 arXiv
[19]

Tomas Mikolov, Ilya Sutskever, Kai Chen, Greg S Corrado, and Jeff Dean. 2013 b . Distributed representations of words and phrases and their compositionality. In Advances in neural information processing systems, volume 26

2013
[20]

Kiho Park, Yo Joong Choe, Yibo Jiang, and Victor Veitch. 2025. https://arxiv.org/abs/2406.01506 The geometry of categorical and hierarchical concepts in large language models . Preprint, arXiv:2406.01506

2025 arXiv
[21]

Kiho Park, Yo Joong Choe, and Victor Veitch. 2024. https://arxiv.org/abs/2311.03658 The linear representation hypothesis and the geometry of large language models . Preprint, arXiv:2311.03658

2024 arXiv
[22]

Ajay Patel, Delip Rao, Ansh Kothary, Kathleen McKeown, and Chris Callison-Burch. 2023. https://doi.org/10.18653/v1/2023.findings-emnlp.1020 Learning interpretable style embeddings via prompting LLM s . In Findings of the Association for Computational Linguistics: EMNLP 2023, p...

2023 doi
[23]

Raphaël Sarfati, Toni J. B. Liu, Nicolas Boullé, and Christopher J. Earls. 2025. https://arxiv.org/abs/2410.01545 Lines of thought in large language models . Preprint, arXiv:2410.01545

2025 arXiv
[24]

William Shakespeare. ca. 1599. https://www.folger.edu/explore/shakespeares-works/romeo-and-juliet/ The Tragedy of Romeo and Juliet . Folger Shakespeare Library. Accessed: 2025-04-07

2025
[25]

Adi Simhi and Shaul Markovitch. 2023. https://openreview.net/forum?id=sPpft5DQJN Interpreting embedding spaces by conceptualization . In The 2023 Conference on Empirical Methods in Natural Language Processing

2023
[26]

Gemma Team, Aishwarya Kamath, Johan Ferret, Shreya Pathak, Nino Vieillard, Ramona Merhej, Sarah Perrin, Tatiana Matejovicova, Alexandre Ramé, Morgane Rivière, Louis Rouillard, Thomas Mesnard, Geoffrey Cideron, Jean bastien Grill, Sabela Ramos, Edouard Yvinec, Michelle Casbon, ...

2025 arXiv
[27]

Daniel Freeman, Theodore R

Adly Templeton, Tom Conerly, Jonathan Marcus, Jack Lindsey, Trenton Bricken, Brian Chen, Adam Pearce, Craig Citro, Emmanuel Ameisen, Andy Jones, Hoagy Cunningham, Nicholas L Turner, Callum McDougall, Monte MacDiarmid, C. Daniel Freeman, Theodore R. Sumers, Edward Rees, Joshua ...

2024
[28]

Tirvengadum

V. Tirvengadum. 1996. https://doi.org/10.16995/dscn.187 Linguistic fingerprints and literary fraud . Digital Studies/le Champ Numérique, 6

1996 doi
[29]

Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timoth \'e e Lacroix, Baptiste Rozi \`e re, Naman Goyal, Eric Hambro, Faisal Azhar, and 1 others. 2023. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971

2023 arXiv
[30]

Lucrezia Valeriani, Diego Doimo, Francesca Cuturello, Alessandro Laio, Alessio Ansuini, and Alberto Cazzaniga. 2023. https://arxiv.org/abs/2302.00294 The geometry of hidden representations of large transformer models . Preprint, arXiv:2302.00294

2023 arXiv
[31]

Karthik Viswanathan, Yuri Gardinazzi, Giada Panerai, Alberto Cazzaniga, and Matteo Biagetti. 2025. https://arxiv.org/abs/2501.10573 The geometry of tokens in internal representations of large language models . Preprint, arXiv:2501.10573

2025 arXiv
[32]

An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, Chujie Zheng, Dayiheng Liu, Fan Zhou, Fei Huang, Feng Hu, Hao Ge, Haoran Wei, Huan Lin, Jialong Tang, and 41 others. 2025. https://arxiv.org/abs/2505.09388 Qw...

2025 arXiv

Pith tools

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

[1] [1]

online" 'onlinestring :=

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRING...

[2] [2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

[3] [3]

Loubna Ben Allal, Anton Lozhkov, Elie Bakouch, Gabriel Martín Blázquez, Guilherme Penedo, Lewis Tunstall, Andrés Marafioti, Hynek Kydlíček, Agustín Piqueres Lajarín, Vaibhav Srivastav, Joshua Lochner, Caleb Fahlgren, Xuan-Son Nguyen, Clémentine Fourrier, Ben Burtenshaw, Hugo Larcher, Haojun Zhao, Cyril Zakka, Mathieu Morlon, and 3 others. 2025. https://ar...

arXiv 2025

[4] [4]

Milad Alshomary, Narutatsu Ri, Marianna Apidianaki, Ajay Patel, Smaranda Muresan, and Kathleen McKeown. 2024. https://arxiv.org/abs/2409.07072 Latent space interpretation for stylistic analysis and explainable authorship attribution . Preprint, arXiv:2409.07072

work page Pith review arXiv 2024

[5] [5]

Trenton Bricken, Adly Templeton, Joshua Batson, Brian Chen, Adam Jermyn, Tom Conerly, Nick Turner, Cem Anil, Carson Denison, Amanda Askell, Robert Lasenby, Yifan Wu, Shauna Kravec, Nicholas Schiefer, Tim Maxwell, Nicholas Joseph, Zac Hatfield-Dodds, Alex Tamkin, Karina Nguyen, and 6 others. 2023. https://transformer-circuits.pub/2023/monosemantic-features...

work page 2023

[6] [6]

Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, and 1 others. 2020. Language models are few-shot learners. In Advances in neural information processing systems, volume 33, pages 1877--1901

work page 2020

[7] [7]

Chang, Zhuowen Tu, and Benjamin K

Tyler A. Chang, Zhuowen Tu, and Benjamin K. Bergen. 2022. https://arxiv.org/abs/2205.10964 The geometry of multilingual language model representations . Preprint, arXiv:2205.10964

arXiv 2022

[8] [8]

Yamshchikov

Svetlana Gorovaia, Gleb Schmidt, and Ivan P. Yamshchikov. 2024. https://doi.org/10.18653/v1/2024.nlp4dh-1.39 S ui generis: Large language models for authorship attribution and verification in L atin . In Proceedings of the 4th International Conference on Natural Language Processing for Digital Humanities, pages 398--412, Miami, USA. Association for Comput...

work page doi:10.18653/v1/2024.nlp4dh-1.39 2024