Under the Shadow of Babel: How Language Shapes Reasoning in LLMs

Chenxi Wang; Lang Gao; Xiuying Chen; Yanbo Wang; Yixuan Zhang; Zirui Song; Zixiang Xu

REVIEW 4 major objections 5 minor 1 cited by

Under the Shadow of Babel: How Language Shapes Reasoning in LLMs

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Large language models internalize language-specific causal habits — attending to sentence-initial causes in Chinese — and the habit degrades reasoning when Chinese input presents effects first.

desk verdict New dataset and attention-analysis lens, but the headline claim about internalized causal priors rests on an unmeasured corpus-frequency assumption and thin statistics. read the letter →

arxiv 2506.16151 v1 pith:U4GR4NN2 submitted 2025-06-19 cs.CL cs.AI

Chenxi Wang , Yixuan Zhang , Lang Gao , Zixiang Xu , Zirui Song , Yanbo Wang , Xiuying Chen This is my paper

classification cs.CLcs.AI

keywords linguisticrelativitycausalreasoninglargelanguagemodelsattentionallocationcross-lingualrepresentationsChinese-EnglishtypologymultilingualinterpretabilityBICAUSE

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 tests whether large language models inherit the reasoning habits of the languages they are trained on — a machine analogue of the linguistic-relativity idea that language shapes thought. To do this it introduces BICAUSE, 400 semantically aligned Chinese and English three-step causal chains, each written in a forward (cause-first) and a reversed (effect-first) order with an inference question. The authors find that models attend differently in the two languages, favor causes and sentence-initial connectives in Chinese, apply this cause-first habit so rigidly that reversed Chinese chains score 76.5% accuracy versus 88.5% for reversed English, and yet converge on nearly identical hidden representations whenever the reasoning succeeds. If the findings hold, they give the first internal-structure evidence that LLMs encode language-specific causal conventions as stable attention strategies, with measurable behavioral cost when input departs from the convention.

What carries the argument

Two instruments carry the argument. BICAUSE is a dataset of 400 semantically and syntactically aligned three-step causal chains in Chinese and English, each in forward and reversed order with a QA inference question, decomposed into 13 syntactic components and 3 causal components ([cause], [intermediate effect], [final effect]). The Relative Component Attention Ratio (RCAR) measures the attention each component receives from later tokens, normalized per token so comparisons across languages are fair; SVCCA (Singular Vector Canonical Correlation Analysis) scores the similarity of the 24-layer attention trajectories across the four input variants, and layerwise cosine similarity of hidden vectors at the final causal token tracks whether successful reasoning shares a representation. The forward/reversed contrast is what makes the internalized prior visible: identical semantics, identical model, different word order, and the attention pattern moves exactly as a typological habit would predict.

What would settle it

Count effect-first causal constructions (with connectives such as '是由于...而这源自') in the pretraining corpora of the tested models and compare the rates with their English counterparts; if reversed Chinese forms occur at comparable or higher frequency than English 'X happened, because Y' forms and the 76.5% versus 88.5% gap persists, the internalized-rarity explanation fails. As a second check, have native Chinese readers rate the naturalness of the reversed BICAUSE sentences: ratings at the level of translationese would indicate the stimuli, not a learned prior, drive the drop.

Watch

Extended reading notes

Core claim

The paper's central claim is that LLMs do not merely process surface syntax: they internalize language-specific habitual patterns of causal expression as stable attention-allocation strategies, and those habits shape downstream reasoning. Because Chinese prefers forward causal order and topic-initial structure, the model learns a 'sentence-initial component is the cause' prior and locks attention onto the first event; applied to a grammatically valid but atypical reversed chain, the prior misfires, the model anchors on the final effect, and accuracy falls to 76.5%. Because English routinely uses effect-first constructions such as 'X happened, because Y', the same model keeps attention flexible and accuracy high (88.5%) under both orders. The paper further argues that the destination of reasoning is language-agnostic: correctly reasoned samples from all four conditions converge on nearly identical hidden representations, so the divergence is in the habitual path, not in the shared understanding reached at the end.

Load-bearing premise

The central premise is that effect-first causal sentences are genuinely rare in the Chinese text the models were trained on; if they are actually common, or if the authors' reversed Chinese sentences read as unnatural templates rather than native usage, the accuracy gap would reflect the test items rather than an internalized cause-first habit.

Editorial extensions

If this is right

Chinese reversed causal chains will keep underperforming English reversed chains across model families and sizes, because the cause-first habit is a property of the pretraining distribution, not of one architecture.
Answer accuracy on a causal question is predictable from attention placement: the SVCCA similarity between a structure and the canonical forward form tracks accuracy, from 0.73 with roughly 91% down to 0.46 with 76.5%.
Multilingual benchmarks that test only canonical word order will overstate cross-lingual reasoning ability; atypical constructions expose the language-specific ceiling.
Because successful reasoning converges on a shared abstraction, cross-lingual semantic alignment is attainable in representation space even while surface processing remains language-specific.

Reading between the lines

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

The same measurement could be run on language pairs with other typological contrasts — head-final languages, for example — to predict which constructions will degrade and which will transfer easily.
A causal test of the attention-as-mechanism reading: redirect attention toward the true cause in reversed Chinese chains by activation patching and check whether accuracy recovers to the English level; if attention is only a correlate, the intervention should do nothing.
The rarity premise is directly checkable: corpus statistics on effect-first causal constructions in native Chinese, plus native-speaker naturalness ratings for the reversed BICAUSE sentences, would confirm or refute the claim that the drop comes from an underrepresented pattern rather than from awkward stimuli.
If the habit is learned from corpus statistics, fine-tuning on reversed Chinese causal chains should shrink the 76.5% versus 88.5% gap — a cheap intervention test with practical value for aligning multilingual models.

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.

Desk Editor's Note

read the letter

The useful core is BICAUSE and the RCAR decomposition. The controlled forward/reversed causal chains in Chinese and English are a good testbed, and the attention divergence between the two languages is a real descriptive finding. The reversed-Chinese accuracy drop on Qwen1.5-1.8B (76.5% vs 88.5%) is concrete. But the paper's central interpretation—that LLMs internalize a language-specific cause-first prior—rests on an assumption that is asserted, not measured. Section 5.2 says reversed Chinese forms are 'rare in native usage and likely underrepresented in the training corpus'; Figure 1 says the same. No corpus counts, no frequency statistics, no human naturalness ratings. If the reversed Chinese template is simply stilted or a calque, the accuracy gap is a stimulus artifact, not evidence about internalized reasoning biases. The stress-test note lands.

What is genuinely new: BICAUSE is carefully aligned semantically and syntactically, and the RCAR metric is clearly defined and reasonable for component-level attention. Running eight models is better than most papers do. The finding that attention tracks word order and connective position across languages is worth having.

The soft spots are real. Four hundred samples total, no confidence intervals, no significance tests; the 12-point gap on the 1.8B model could easily be noise. More seriously, the effect does not generalize across the appendix models: Qwen1.5-32B gets 94.0 on Chinese reversed versus 90.5 on English reversed; 14B and 72B show almost no gap. The text says the conclusion is generalizable, but the tables contradict it. The representation-convergence analysis (Figure 5) is restricted to samples where both inputs are correct, which biases the cosine similarities upward and makes the convergence claim close to circular. The typological explanation is plausible but post hoc; it is not tested against the possibility that the specific templates, or tokenization, drive the differences.

This paper is for people working on multilingual interpretability and cross-lingual evaluation. The dataset and analysis framework deserve engagement, but the Whorfian claim outruns the evidence. A serious referee should send it back for significance testing, corpus frequency counts for the reversed Chinese constructions, naturalness ratings, and an honest treatment of the model-size inconsistency. I would not desk-reject it, and I'd be willing to referee it myself, but I wouldn't cite the internalization claim in its current form.

Referee Report

4 major / 5 minor

Summary. The paper introduces BICAUSE, a bilingual Chinese-English causal reasoning dataset with semantically aligned forward and reversed causal chains across eight domains. Using the RCAR attention metric (Eq. 1-2), SVCCA, and hidden-state cosine similarities on Qwen1.5-1.8B-Chat plus several larger models, it reports three findings: LLMs allocate attention in typologically aligned ways (more subject/connective focus in Chinese, more verb/result focus in English); they internalize a cause-first ordering preference for Chinese that degrades accuracy on reversed Chinese chains (Table 2: Qwen1.5-1.8B 76.8% Chinese vs 88.5% English); and successful reasoning is accompanied by cross-language convergence of hidden representations. The authors interpret these results as evidence that LLMs internalize language-specific causal reasoning biases, echoing linguistic relativity.

Significance. If the claims hold, the paper provides a new kind of evidence for linguistic relativity in LLMs and a reusable dataset and analytical framework for cross-lingual interpretability. The dataset is publicly released, the RCAR metric is explicitly defined, and the main attention analyses are replicated across several model families and sizes, which are concrete strengths. The central limitation is that the behavioral asymmetry is attributed to an unmeasured corpus-frequency asymmetry, so the current evidence does not yet uniquely support the linguistic-relativity reading. The representation-convergence claim is also weakened by selection on correct predictions. These issues are addressable, however, and the paper's overall direction is sound.

major comments (4)

[Section 5.2, Figure 1, Table 2] The claim that reversed causal chains are 'rare in native usage and likely underrepresented in the training corpus' is the load-bearing premise for explaining the Chinese reversed-chain accuracy drop as an internalized cause-first prior. The manuscript provides no corpus counts, n-gram frequencies, independent citations, or human naturalness ratings for the reversed Chinese template such as '香气扩散，是由于面包烤熟，而这源自面包机加热'. Without such measurements, the 12-point accuracy gap in Table 2 could be a stimulus-construction artifact, e.g., if the Chinese reversed template is a formal or literary calque of the English 'due to ... which originates from'. I request a concrete test: report corpus frequencies of the forward and reversed construction types in large Chinese and English corpora, and/or collect native-speaker naturalness ratings, and show that model accuracy tracks this measured asymmetry.
[Section 4.2-4.3, Figures 2-3, Tables 1-2] The central quantitative comparisons, including RCAR differences up to 1.585 in Figure 3 and the Chinese reversed accuracy drop, are presented without confidence intervals, bootstrap estimates, or significance tests, despite the dataset containing only 400 samples (50 per domain). For instance, Table 2 reports Qwen1.5-1.8B at 88.5% (English) versus 76.8% (Chinese) on reversed chains, but no uncertainty is attached and the domain-level values show high variance (Qwen1.5-1.8B Chinese ranges from 64% to 86% across domains). Please report per-condition confidence intervals and pairwise tests across models before claiming a systematic Chinese-specific degradation.
[Section 6, Figure 5] The claim that successful reasoning converges to shared, language-agnostic representations is based on cosine similarity computed only over pairs where both predictions are correct. This selection on the dependent variable can inflate similarity, because correct cases may be systematically easier or less noisy, so the convergence could reflect task difficulty rather than language-agnostic abstraction. Please report the same analysis on all samples or on matched correct/incorrect subsets, and compare against a shuffled-token baseline, to show that the convergence is not an artifact of filtering.
[Section 4.3] The typological interpretation of attention divergence is plausible but confounded with surface word order and the connective inventory. For example, 'once' appears sentence-initially in both languages in the forward template, while Chinese and English connectives have different positional and co-occurrence distributions. A stronger test would include English stimuli with Chinese-like preposed connectives or Chinese stimuli with English-like postverbal connectives, to separate typological bias from template-specific surface effects.

minor comments (5)

[Section 5.2 and Table 2] The text reports the Chinese reversed-chain accuracy as 76.5%, but Table 2 lists 76.8%; please reconcile the numbers.
[Section 3 and Appendix A] Table 2 is labeled 'paraphrased data,' while Section 3 consistently calls these examples 'reversed causal chains'; please unify the terminology and clarify whether the 400-sample total refers to forward chains, reversed chains, or chain semantics shared across both forms.
[Figure 4 caption] The caption contains a typo, 'LLayerwise,' and the figure label in Section 5.1 also says 'lanuages'; please correct these.
[References] Several reference entries are malformed or inconsistent, including 'Steven Piantadosi' as the author of 'Meaning without reference in large language models' (should be Piantadosi and Hill), the formatting of the Feng-Fu citation, and 'V oita' in the Voita et al. entry; please check the bibliography.
[Figure 3] The bars appear sorted by magnitude, but the sorting criterion is not stated; please describe the ordering and add error bars or a significance annotation.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the empirical measurements (RCAR, SVCCA, accuracy) are made on an externally constructed dataset, and the typological explanation is an external hypothesis, not an input to the derivation.

full rationale

No significant circularity is present. The paper's empirical chain—constructing the BICAUSE dataset, computing RCAR attention ratios (Eq. 1-2), SVCCA similarities, and QA accuracy—is performed on an externally constructed dataset, and no parameter is fitted to the target conclusion. The observation that Chinese reversed causal chains receive different attention and lower accuracy (76.5% vs. 88.5% English reversed) is a direct measurement, not an identity derived from the RCAR definition, and no equation in the paper reduces to its own inputs. The typological interpretation in Sections 4.3 and 5.2 (Chinese topic-prominence, cause-first preference, rarity of reversed Chinese forms) is an external linguistic hypothesis used to explain the measurements, not a premise used to construct them; thus there is no Eq. X = Eq. Y reduction. The main weakness is the unsupported empirical claim in Section 5.2 that reversed causal forms 'are rare in native usage and likely underrepresented in the training corpus,' which is asserted without corpus frequency counts or native-speaker naturalness ratings. If false, the Chinese accuracy drop could be a stimulus artifact rather than evidence of an internalized typological prior, but this is an empirical-validity and correctness concern, not circularity. The paper cites prior work (e.g., Schut et al. 2025, Lindsey et al. 2025) only as convergent support for representation alignment findings, and no load-bearing argument reduces to a self-citation chain.

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

The paper introduces no fitted free parameters and no invented entities. Its conclusions rest on four assumptions: that reversed Chinese causal forms are rare in pretraining data (unmeasured), that normalized attention ratios are comparable across tokenizers, that the final causal-chain token captures the integrated causal state, and that SVCCA and cosine similarity summarize the relevant structure. These are reasonable but not independently verified within the paper.

assumptions (4)

domain assumption Chinese reversed causal chains are rare in native usage and underrepresented in pretraining corpora
Invoked in Section 5.2 to explain the Chinese reversed-chain accuracy drop; no corpus counts or independent frequency evidence are provided.
domain assumption Two-level normalization makes component-level attention ratios comparable across Chinese and English despite different tokenizers and segmentation granularity
Section 4.1 relies on RCAR comparability; tokenizer differences are not analyzed.
domain assumption The hidden state at the final causal-chain token captures the integrated causal representation
Section 6 uses this position for cosine similarity; no control experiment validates the choice.
standard math SVCCA and cosine similarity are valid measures of structural and representational alignment
SVCCA is a standard method (Raghu et al., 2017), but conclusions depend on the 24x3 RCAR trajectories being a sufficient summary of attention.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Under the Shadow of Babel: How Language Shapes Reasoning in LLMs." pith.science (2026). https://pith.science/paper/U4GR4NN2

@misc{pith2026250616151,
  author       = {Pith},
  title        = {Pith review of: Under the Shadow of Babel: How Language Shapes Reasoning in LLMs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/U4GR4NN2}},
  note         = {Machine review of arXiv:2506.16151}
}

read the original abstract

Language is not only a tool for communication but also a medium for human cognition and reasoning. If, as linguistic relativity suggests, the structure of language shapes cognitive patterns, then large language models (LLMs) trained on human language may also internalize the habitual logical structures embedded in different languages. To examine this hypothesis, we introduce BICAUSE, a structured bilingual dataset for causal reasoning, which includes semantically aligned Chinese and English samples in both forward and reversed causal forms. Our study reveals three key findings: (1) LLMs exhibit typologically aligned attention patterns, focusing more on causes and sentence-initial connectives in Chinese, while showing a more balanced distribution in English. (2) Models internalize language-specific preferences for causal word order and often rigidly apply them to atypical inputs, leading to degraded performance, especially in Chinese. (3) When causal reasoning succeeds, model representations converge toward semantically aligned abstractions across languages, indicating a shared understanding beyond surface form. Overall, these results suggest that LLMs not only mimic surface linguistic forms but also internalize the reasoning biases shaped by language. Rooted in cognitive linguistic theory, this phenomenon is for the first time empirically verified through structural analysis of model internals.

Figures

Figures reproduced from arXiv: 2506.16151 by the authors.

**Figure 2.** Layerwise RCAR over eight syntactic components in Chinese (orange) and English (blue) causal chains. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

**Figure 3.** RCAR differences between Chinese and En [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Figures from the paper (7 more)

**Figure 4.** Figure 4: LLayerwise RCAR heatmaps for three core causal components — [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

**Figure 5.** Figure 5: Layerwise cosine similarity of hidden vectors [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

**Figure 6.** Figure 6: Layerwise RCAR over all syntactic components in Chinese (orange) and English (blue) causal chains. [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]

**Figure 7.** Figure 7: Layerwise RCAR heatmaps for Qwen1.5-7B-Chat showing attention patterns across causal components [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]

**Figure 8.** Figure 8: Layerwise RCAR heatmaps for Qwen1.5-14B-Chat. The larger model size shows more structured attention [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]

**Figure 9.** Figure 9: Layerwise RCAR heatmaps for Qwen1.5-32B-Chat. With 32 billion parameters, this model demonstrates [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]

**Figure 10.** Figure 10: Layerwise RCAR heatmaps for Qwen1.5-72B-Chat. As the largest model in the Qwen family, it [PITH_FULL_IMAGE:figures/full_fig_p015_10.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

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

Inside the Unfair Judge: A Mechanistic Interpretability Account of LLM-as-Judge Bias
cs.LG 2026-07 conditional novelty 6.5 of 10

LLM-as-judge scoring biases concentrate in low-dimensional, type-specific activation subspaces that support bidirectional causal steering and cross-domain failure prediction.

Reference graph

Works this paper leans on

15 extracted references · 11 canonical work pages · cited by 1 Pith paper

[1]

2011.Badan L

Linda Badan and Francesca Gobbo. 2011.Badan L. e Del Gobbo F

work page 2011
[7]

InProceed- ings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 12216–12235

Dissecting recall of factual associa- tions in auto-regressive language models. InProceed- ings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 12216–12235. M.A.K. Halliday and Christian Matthiessen. 2004.An Introduction to Functional Grammar. Johann Gottfried Herder and Jean-Jacques Rousseau. 1986.On the Origin of Langua...

work page 2023
[8]

Wilhelm Humboldt, editor

A survey on large language models with multilingualism: Recent advances and new frontiers.Preprint, arXiv:2405.10936. Wilhelm Humboldt, editor

arXiv
[9]

https://transformer-circuits.pub/2025/ attribution-graphs/biology.html

On the biology of a large language model. https://transformer-circuits.pub/2025/ attribution-graphs/biology.html. John A. Lucy. 1992.Language Diversity and Thought: A Reformulation of the Linguistic Relativity Hypothe- sis. Cambridge University Press. Kevin Meng, David Bau, Alex Andonian, and Yonatan Belinkov

work page 2025
[10]

Aneta Pavlenko

Decomposed prompting: Unveiling multilin- gual linguistic structure knowledge in english-centric large language models.Preprint, arXiv:2402.18397. Aneta Pavlenko

arXiv
[11]

John Schwieter

Do multilingual llms think in english?Preprint, arXiv:2502.15603. John Schwieter

arXiv
[13]

SeaE- val for multilingual foundation models: From cross- lingual alignment to cultural reasoning. InProceed- ings of the 2024 Conference of the North American Chapter of the Association for Computational Lin- guistics: Human Language Technologies (V olume 1: Long Papers), pages 370–390, Mexico City, Mexico. Association for Computational Linguistics. Chri...

work page 2024
[14]

InProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (V olume 1: Long Papers), pages 15366–15394, Bangkok, Thai- land

Do llamas work in English? on the latent language of multilingual transformers. InProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (V olume 1: Long Papers), pages 15366–15394, Bangkok, Thai- land. Association for Computational Linguistics. Benjamin Lee Whorf. 1956.Language, Thought, and Reality: Selected Writings of ...

work page 1956

Show all 15 references

[15]

InFindings of the Association for Computational Linguistics: ACL 2024, pages 9191– 9200, Bangkok, Thailand

AutoCAP: Towards auto- matic cross-lingual alignment planning for zero-shot chain-of-thought. InFindings of the Association for Computational Linguistics: ACL 2024, pages 9191– 9200, Bangkok, Thailand. Association for Computa- tional Linguistics. A Accuracy Results on BICAUSE ...

2024
[2019]

InProceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP, page

What does bert look at? an analysis of bert’s attention. InProceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP, page

2019
[2020]

InProceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 8718–8735, Online

Experience grounds language. InProceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 8718–8735, Online. Association for Computational Linguistics. Lera Boroditsky

2020
[2022]

InNeurIPS 2022 Workshop on Neuro Causal and Symbolic AI (nCSI)

Meaning without reference in large language models. InNeurIPS 2022 Workshop on Neuro Causal and Symbolic AI (nCSI). Qwen Team

2022
[2023]

Qwen technical report.arXiv preprint arXiv:2309.16609. Yonatan Bisk, Ari Holtzman, Jesse Thomason, Jacob Andreas, Yoshua Bengio, Joyce Chai, Mirella Lap- ata, Angeliki Lazaridou, Jonathan May, Aleksandr Nisnevich, Nicolas Pinto, and Joseph Turian

arXiv
[2024]

Association for Computational Lin- guistics

Do mul- tilingual language models think better in English? InProceedings of the 2024 Conference of the North American Chapter of the Association for Computa- tional Linguistics: Human Language Technologies (V olume 2: Short Papers), pages 550–564, Mexico City, Mexico. Associat...

2024
[2025]

Large language models share representations of latent grammatical concepts across typologically diverse languages. InProceed- ings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computa- tional Linguistics: Human Language Technologies (V o...

2025

Pith tools

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

[1] [1]

2011.Badan L

Linda Badan and Francesca Gobbo. 2011.Badan L. e Del Gobbo F

work page 2011

[2] [7]

InProceed- ings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 12216–12235

Dissecting recall of factual associa- tions in auto-regressive language models. InProceed- ings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 12216–12235. M.A.K. Halliday and Christian Matthiessen. 2004.An Introduction to Functional Grammar. Johann Gottfried Herder and Jean-Jacques Rousseau. 1986.On the Origin of Langua...

work page 2023

[3] [8]

Wilhelm Humboldt, editor

A survey on large language models with multilingualism: Recent advances and new frontiers.Preprint, arXiv:2405.10936. Wilhelm Humboldt, editor

arXiv

[4] [9]

https://transformer-circuits.pub/2025/ attribution-graphs/biology.html

On the biology of a large language model. https://transformer-circuits.pub/2025/ attribution-graphs/biology.html. John A. Lucy. 1992.Language Diversity and Thought: A Reformulation of the Linguistic Relativity Hypothe- sis. Cambridge University Press. Kevin Meng, David Bau, Alex Andonian, and Yonatan Belinkov

work page 2025

[5] [10]

Aneta Pavlenko

Decomposed prompting: Unveiling multilin- gual linguistic structure knowledge in english-centric large language models.Preprint, arXiv:2402.18397. Aneta Pavlenko

arXiv

[6] [11]

John Schwieter

Do multilingual llms think in english?Preprint, arXiv:2502.15603. John Schwieter

arXiv

[7] [13]

SeaE- val for multilingual foundation models: From cross- lingual alignment to cultural reasoning. InProceed- ings of the 2024 Conference of the North American Chapter of the Association for Computational Lin- guistics: Human Language Technologies (V olume 1: Long Papers), pages 370–390, Mexico City, Mexico. Association for Computational Linguistics. Chri...

work page 2024

[8] [14]

InProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (V olume 1: Long Papers), pages 15366–15394, Bangkok, Thai- land

Do llamas work in English? on the latent language of multilingual transformers. InProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (V olume 1: Long Papers), pages 15366–15394, Bangkok, Thai- land. Association for Computational Linguistics. Benjamin Lee Whorf. 1956.Language, Thought, and Reality: Selected Writings of ...

work page 1956