KatotohananQA: Evaluating Truthfulness of Large Language Models in Filipino

Lorenzo Alfred Nery; Ronald Dawson Catignas; Thomas James Tiam-Lee

REVIEW 4 major objections 4 minor 17 references

KatotohananQA: Evaluating Truthfulness of Large Language Models in Filipino

T0 review · 4 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read A Filipino translation of TruthfulQA shows LLMs are systematically less truthful in Filipino, averaging an 11-point accuracy gap.

desk verdict A useful Filipino TruthfulQA resource with a plausible but not fully pinned-down English-Filipino gap; the translation-equivalence worry is real but the paper is honest about its limits. read the letter →

arxiv 2509.06065 v1 pith:5IRUUUJJ submitted 2025-09-07 cs.CL

Lorenzo Alfred Nery , Ronald Dawson Catignas , Thomas James Tiam-Lee This is my paper

classification cs.CL

keywords KatotohananQATruthfulFilipinomultilingualevaluationtruthfulnesslargelanguagemodelslow-resourcelanguagesbenchmarktranslation

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 introduces KatotohananQA, a Filipino translation of the 790-question binary-choice TruthfulQA benchmark, and uses it to ask whether large language models are as truthful in Filipino as they are in English. Seven free-tier proprietary models were evaluated in both languages with matched prompts. The central finding is a consistent gap: average accuracy is 94.72% in English but 83.87% in Filipino, and every model except GPT-5 scores lower in Filipino, with the largest drops on questions that require logical reasoning, cultural knowledge, or identifying people. The paper argues the gap is not uniform: some categories and topics lose far more accuracy than others, and the newest GPT-5 models are the most multilingual-robust. If the dataset is a faithful parallel translation, English truthfulness scores materially overstate how truthful these models are for Filipino speakers.

What carries the argument

The central object is KatotohananQA itself, a parallel Filipino version of TruthfulQA's binary-choice setting containing 790 questions in 37 categories and 19 latent topics. Binary-choice means each question is presented with exactly one correct and one incorrect answer of similar length, so a model cannot win by length-based heuristics. The dataset carries the argument because every result is a paired English-versus-Filipino accuracy comparison on the same question instances; the two-step translation protocol — machine translation plus native-speaker revision, with explicit rules for proper nouns, acronyms, and proverbs — is what licenses treating the two versions as measuring the same know

What would settle it

Back-translate the 790 Filipino items to English and have independent native Filipino-English bilingual raters flag every item whose meaning, option difficulty, or cultural reference shifted in translation. If the accuracy gap on flagged items is comparable to the gap on un-flagged items, the translation-parallelism explanation is supported; if flagged items account for most of the errors, the comparison is not parallel and the central claim would need to be re-assessed.

Watch

Extended reading notes

Core claim

KatotohananQA is a parallel Filipino adaptation of the binary-choice TruthfulQA benchmark: 790 questions, each with one correct and one incorrect option of similar length, machine-translated and then revised by native Filipino speakers under guidelines meant to preserve meaning and cultural context. Across seven free-tier proprietary models evaluated zero-shot, English accuracy averaged 94.72% and Filipino accuracy 83.87%, a mean difference of +10.85 percentage points. The gap was positive for six of the seven models, from +1.01 points for GPT-5 mini to +27.46 for DeepSeek V3, and zero for GPT-5. Paired significance tests and effect sizes show the differences are non-trivial for most models,

Load-bearing premise

The translated Filipino items are parallel to the English binary-choice items in meaning and difficulty, so the observed accuracy gap measures language-specific truthfulness rather than translation artifacts.

Editorial extensions

If this is right

If the central claim is right, English-only TruthfulQA scores overstate Filipino truthfulness for most current free-tier models by roughly 11 points on average.
The gap is largest on reasoning-heavy and culturally embedded items, so a model that looks reliable on English logic or trivia should not be assumed reliable for reasoning in Filipino.
Newer GPT-5 and GPT-5 mini show near-zero gaps, suggesting multilingual truthfulness may be improving in the newest training runs, but the improvement is not yet consistent across vendors.
Safety and deployment evaluations in Filipino-speaking contexts should include native-language truthfulness items rather than relying only on translated general benchmarks.
Category- and topic-level results can flag where a model is least trustworthy in Filipino, guiding targeted data collection or training interventions.

Reading between the lines

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

The translation-parallelism assumption is untested: a back-translation or difficulty-equivalence study of the 790 items could change how much of the 11-point gap is attributed to language-specific truthfulness versus translation artifacts.
The concentration of the gap in logic puzzles, riddles, and clue-based identification suggests the bottleneck may be Filipino-mediated reasoning rather than factual coverage; fine-tuning on Filipino reasoning data would be a targeted probe of that hypothesis.
Because the new benchmark is public, future model releases can be checked against it directly, turning the reported gap into a trackable metric that does not require re-translating the questions.
A natural extension is to measure whether the binary-choice gap persists in free-form generation, where a model's tendency to hedge, switch languages, or produce false explanations may be even more visible.

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

Read this if you're working on multilingual evaluation or truthfulness benchmarks. The dataset is the contribution: a 790-item Filipino translation of binary-choice TruthfulQA, human-verified by native speakers, with clear translation guidelines. That's useful. The evaluation of seven proprietary models is straightforward and the results are reported per model, type, category, and topic, with McNemar and Cohen's g. The headline gap—English 94.72, Filipino 83.87—is large and the discordant pairs (650/70, though Table 8 says 670) back it up.

The soft spots are not fatal but they matter. The translation-equivalence worry is real: no back-translation, no inter-annotator agreement, and the examples show intended code-mixing where English phrases are kept for cultural fidelity. That could make Filipino items harder or easier in ways that aren't measured. The paper doesn't claim to have solved that, and it lists limitations, but the central interpretation—that the gap reflects training-data disparity—would be stronger with a difficulty check. Also, the OpenAI robustness claim rests on two models, single runs, temperature 1.0, and no contamination check; that's a thin reed for a "strong multilingual robustness" conclusion. And the 650/670 mismatch in the contingency table should be fixed.

Credit where due: the dataset is publicly available, the method is clearly described, and the statistical reporting is better than most benchmark papers. The related work is appropriately cited, including Uhura's warning that translations may not be perfectly parallel—so the authors were aware of the risk.

Who is it for? People building multilingual benchmarks and anyone studying cross-lingual truthfulness. It deserves a serious referee, with requests for the equivalence analysis and corrected numbers. My recommendation: send it to review, conditionally. It's a solid resource with an honest limitations section, not a paper that needs to be reshaped.

Referee Report

4 major / 4 minor

Summary. The paper introduces KatotohananQA, a Filipino adaptation of the 790-item binary-choice TruthfulQA benchmark, built by machine-translating with Google Translate and then revising by native speakers. The authors evaluate seven free-tier proprietary LLMs in English and Filipino, reporting an overall accuracy of 94.72% in English versus 83.87% in Filipino, a mean gap of +10.85 percentage points. They further break down performance by model, question type, category, and topic, and use McNemar's test and Cohen's g to assess significance and effect size. The paper concludes that LLMs are less truthful in Filipino than in English, that GPT-5 and GPT-5 mini are relatively robust multilingually, and that certain question categories and topics are especially sensitive to language transfer.

Significance. KatotohananQA addresses a real gap: truthfulness evaluation in low-resource languages is scarce, and the binary-choice TruthfulQA format is a useful testbed. The dataset is publicly released, which is a concrete contribution to multilingual evaluation. If the reported gap is real, the finding that English accuracy systematically overstates Filipino truthfulness for several proprietary models is important for deployment and fairness. The paper also usefully connects to prior multilingual TruthfulQA work. However, the central quantitative claim depends on the translated items being parallel to the English items in meaning and difficulty, and that parallelism is not demonstrated. As it stands, the paper establishes an association between language and accuracy, but not yet that the gap is a language-specific truthfulness deficit rather than a translation artifact.

major comments (4)

[§3.1, Tables 1–3] The headline claim that English truthfulness overstates Filipino truthfulness assumes that KatotohananQA items are parallel to the English binary-choice TruthfulQA items in meaning and difficulty. The paper reports a two-step translation process but provides no back-translation check, no difficulty-equivalence analysis, and no inter-annotator agreement metrics. The examples in Tables 1–3 show that many items deliberately retain English words or whole phrases (e.g., 'Council Housed and Violent', 'eating an apple a day'), producing code-mixed prompts. If these prompts are less natural, more ambiguous, or systematically harder to parse for the models than the English originals, the observed overall gap (−10.85 points, Table 4) and the category- and topic-level gaps could be artifacts of the adaptation. Section 5.3 attributes the gap to training-data disparity, but this confound is not ruled
[§3.3, §3.5] Each model was run once per item, yet statistical significance is claimed and effect sizes are reported. Moreover, GPT-5 and GPT-5 mini were evaluated at temperature 1.0 while all other models used temperature 0.0. With a single stochastic sample at temperature 1.0, the reported 0.00-point gap for GPT-5 and the 1.01-point gap for GPT-5 mini could be partly sampling noise. The paper should either match decoding settings across models, report repeated sampling with confidence intervals, or at least quantify the variance of accuracy under the actual sampling scheme. This is directly relevant to the 'multilingual robustness of OpenAI models' conclusion in §5.2.
[§5.1 vs Table 8] The text states that across 5,530 instances, English was correct and Filipino incorrect 650 times, with 70 of the reverse pattern. Table 8 reports 670 in the English-correct/Filipino-incorrect cell. The table is internally consistent with the reported overall accuracies (5238/5530 ≈ 94.72%; 4638/5530 ≈ 83.87%), so the number 650 appears to be a typographical error, but the inconsistency must be corrected. If 670 is correct, the point is actually strengthened, but the text and table should agree.
[§3.5, §4] The paper does not report the actual McNemar p-values or Cohen's g values for the individual models, categories, or topics; it only states how many groups were significant. Without these values, the reader cannot assess the strength of evidence, and no correction for multiple comparisons is mentioned. Given that 37 categories and 19 topics are tested, some significant results would be expected by chance. Please provide full statistics in a supplementary table or appendix.

minor comments (4)

[Tables 6 and 7] The 'Average' rows in these tables are unweighted means over the category/topic rows, not pooled accuracies across all items. This should be stated in the table caption or text to avoid confusion with the overall accuracy in Table 4.
[§5.3] The sentence 'These findings suggest the performance gap between English and Filipino isn’t about a lack of knowledge, but a disparity in training data' overstates what the data can show. The benchmark does not directly measure training data proportions or knowledge separation; this is a plausible hypothesis but should be framed as such.
[§5.2] The GPT-5 and GPT-5 mini results (97.72% and 93.67% in Filipino) are near the English ceiling. The interpretation that this reflects 'multilingual robustness' should acknowledge possible ceiling effects and the fact that temperature settings differ.
[General] Minor typographical/style issues: 'GPT-5 Mini' and 'GPT-5 mini' are used inconsistently; Table 7 lists a topic called 'Language' with different values from the category 'Language' in Table 6, which may confuse readers; and the reference to the OpenAI blog [17] could be supplemented by a more detailed technical source.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper is an empirical benchmark evaluation against an external standard, not a derivation from fitted inputs or self-citations.

full rationale

KatotohananQA is a Filipino translation of the external TruthfulQA benchmark, and the paper's central claims are empirical comparisons of model accuracies on the original English items versus the translated Filipino items. There is no fitted parameter renamed as a prediction, no quantity defined in terms of the outcome it is claimed to predict, and no self-citation chain that forces the reported results. The references to related Filipino benchmarks (e.g., Batayan, FilBench) are contextual and not load-bearing for the accuracy measurements. Concerns about translation quality, code-mixing, and lack of back-translation or difficulty-equivalence checks are validity/threats-to-inference issues, not circularity. The reported McNemar and Cohen's g analyses are standard statistical summaries of the collected responses, and the paper's limitations section acknowledges the scope restrictions. Thus the derivation chain is self-contained with respect to circularity.

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

The central empirical claims rest on benchmark-validity assumptions rather than fitted parameters. No numbers are fit to data; the reported accuracies are measured outcomes. The main risk is that translation changes difficulty or labels, making the language gap partly a test artifact.

assumptions (5)

domain assumption The machine-translated and native-speaker-revised Filipino items preserve the meaning and difficulty of the English TruthfulQA binary-choice items.
Section 3.1 describes the translation workflow but provides no back-translation, difficulty-equivalence check, or inter-annotator agreement.
domain assumption The English TruthfulQA binary-choice answers and labels remain valid ground truth for the Filipino version.
The paper keeps Western cultural contexts and English proper nouns, but does not revalidate correctness labels in the Filipino context.
domain assumption Matching the first letter of the model response to the answer label is a valid proxy for choosing an option.
Section 3.4 says explanations are tolerated and only the first letter is parsed, assuming models still reveal their selected option.
domain assumption The English and Filipino prompts are parallel except for language, so performance differences reflect language capability.
Listing 1.1 shows matched prompt structures, but GPT-5 and GPT-5 mini were run at temperature 1.0 while others ran at 0.0.
domain assumption GPT-5 and GPT-5 mini did not differentially benefit from prior exposure to TruthfulQA training data.
Section 5.2 attributes OpenAI robustness to newer release and training diversity without testing for contamination by the public TruthfulQA benchmark.

how reviews work

0 comments

Cite this review

Pith. "Pith review of KatotohananQA: Evaluating Truthfulness of Large Language Models in Filipino." pith.science (2026). https://pith.science/paper/5IRUUUJJ

@misc{pith2026250906065,
  author       = {Pith},
  title        = {Pith review of: KatotohananQA: Evaluating Truthfulness of Large Language Models in Filipino},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5IRUUUJJ}},
  note         = {Machine review of arXiv:2509.06065}
}

read the original abstract

Large Language Models (LLMs) achieve remarkable performance across various tasks, but their tendency to produce hallucinations limits reliable adoption. Benchmarks such as TruthfulQA have been developed to measure truthfulness, yet they are primarily available in English, leaving a gap in evaluating LLMs in low-resource languages. To address this, we present KatotohananQA, a Filipino translation of the TruthfulQA benchmark. Seven free-tier proprietary models were assessed using a binary-choice framework. Findings show a significant performance gap between English and Filipino truthfulness, with newer OpenAI models (GPT-5 and GPT-5 mini) demonstrating strong multilingual robustness. Results also reveal disparities across question characteristics, suggesting that some question types, categories, and topics are less robust to multilingual transfer which highlight the need for broader multilingual evaluation to ensure fairness and reliability in LLM usage.

Figures

Figures reproduced from arXiv: 2509.06065 by the authors.

**Figure 1.** Effect Size Distribution per Group To assess whether these differences were statistically significant and to quantify their magnitude, McNemar’s test and Cohen’s g were applied both overall and within each grouping. Results showed significant differences for 5/7 of models, 2/2 of question types, 26/37 of question categories, and 19/19 question topics. The distribution of the effect sizes are shown in [PITH_FULL_I… view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

17 extracted references · 12 canonical work pages

[1]

https: //www.nature.com/articles/s41598-025-98483-1

Raza, M., Jahangir, Z., Riaz, M.B., Saeed, M.J., Sattar, M.A.: Industrial ap- plications of large language models.Scientific Reports15, 98483 (2025). https: //www.nature.com/articles/s41598-025-98483-1

work page 2025
[2]

https: //www.nature.com/articles/s41586-024-07421-0

Farquhar, S., Kossen, J., Kuhn, L., Gal, Y.: Detecting hallucinations in large lan- guage models using semantic entropy.Nature630(8017), 625–630 (2024). https: //www.nature.com/articles/s41586-024-07421-0

work page 2024
[3]

https://arxiv.org/abs/2311.05232

Huang, L., Yu, W., Ma, W., Zhong, W., Feng, Z., Wang, H., Chen, Q., Peng, W., Feng, X., Qin, B., Liu, T.: A survey on hallucination in large language models: prin- ciples, taxonomy, challenges, and open questions.arXiv preprint arXiv:2311.05232 (2024). https://arxiv.org/abs/2311.05232

arXiv 2024
[4]

https://arxiv.org/abs/2109

Lin, S., Hilton, J., Evans, O.: TruthfulQA: Measuring how models mimic human falsehoods.arXiv preprint arXiv:2109.07958(2022). https://arxiv.org/abs/2109. 07958

arXiv 2022
[5]

LessWrong(2025)

Evans, O., Chua, J., Lin, S.: New, improved multiple-choice TruthfulQA. LessWrong(2025). https://www.lesswrong.com/posts/Bunfwz6JsNd44kgLT/ new-improved-multiple-choice-truthfulqa 14 Nery et al

work page 2025
[6]

Association for Computational Lin- guistics

¨Ust¨ un, A., Aryabumi, V., Yong, Z., Ko, W., D’souza, D., Onilude, G., Bhandari, N., Singh, S., Ooi, H.-L., Kayid, A., Vargus, F., Blunsom, P., Longpre, S., Muennighoff, N., Fadaee, M., Kreutzer, J., Hooker, S.: Aya Model: An Instruction Finetuned Open-Access Multilingual Language Model.Proceedings of the 62nd Annual Meet- ing of the Association for Comp...

work page 2024
[7]

https://commoncrawl.github.io/ cc-crawl-statistics/plots/languages

Common Crawl: Statistics of Common Crawl Monthly Archives – Lan- guages.Common Crawl GitHub Pages(2025). https://commoncrawl.github.io/ cc-crawl-statistics/plots/languages

work page 2025
[8]

https://arxiv.org/abs/2310.16787

Longpre, S., Mahari, R., Chen, A., Obeng-Marnu, N., Sileo, D., Brannon, W., Muennighoff, N., Khazam, N., Kabbara, J., Perisetla, K., Wu, X., Shippole, E., Bollacker, K., Wu, T., Villa, L., Pentland, S., Hooker, S.: The Data Provenance Initiative: A Large Scale Audit of Dataset Licensing & Attribution in AI.arXiv preprint arXiv:2310.16787(2023). https://ar...

arXiv 2023

Show all 17 references

[9]

R., Layacan, J

Montalan, J. R., Layacan, J. P., Africa, D. D., Flores, R. I., Lopez II, M. T., Magsajo, T. D., Cayabyab, A., Tjhi, W. C.: Batayan: A Filipino NLP bench- mark for evaluating large language models.arXiv preprint arXiv:2502.14911(2025). https://arxiv.org/abs/2502.14911

2025 arXiv
[10]

arXiv preprint arXiv:2502.09387(2025)

Figueras, B.C., Sagarzazu, E., Etxaniz, J., Barnes, J., Gamallo, P., De Dios Flores, I., Agerri, R.: Truth Knows No Language: Evaluating Truthfulness Beyond English. arXiv preprint arXiv:2502.09387(2025). https://arxiv.org/abs/2502.09387

2025
[11]

Association for Computational Linguistics

Aula-Blasco, J., Falc˜ ao, J., Sotelo, S., Paniagua, S., Gonzalez-Agirre, A., Ville- gas, M.: VeritasQA: A Truthfulness Benchmark Aimed at Multilingual Transferabil- ity.Proceedings of the 31st International Conference on Computational Linguistics, pages 5463–5474, Abu Dhabi, ...

2025
[12]

A., Alabi, J

Bayes, E., Azime, I. A., Alabi, J. O., Kgomo, J., Eloundou, T., Proehl, E., Chen, K., Khadir, I., Etori, N. A., Muhammad, S. H., Mpanza, C., Thete, I. P., Klakow, D., Adelani, D. I.: Uhura: A Benchmark for Evaluating Scientific Question Answering and Truthfulness in Low-Resour...

2024 arXiv
[13]

https://arxiv.org/abs/2203

Grootendorst, M.: BERTopic: Neural topic modeling with a class-based TF-IDF procedure.arXiv preprint arXiv:2203.05794(2022). https://arxiv.org/abs/2203. 05794

2022 arXiv
[14]

S.: Chapter 5: Tests for Paired Nominal Data.R Companion Hand- book(2025)

Mangiafico, S. S.: Chapter 5: Tests for Paired Nominal Data.R Companion Hand- book(2025). https://rcompanion.org/handbook/H 05.html

2025
[15]

https://arxiv.org/abs/2508.03523

Miranda, L.J.V., Aco, E., Manuel, C., Cruz, J.C.B., Imperial, J.M.: FilBench: Can LLMs Understand and Generate Filipino?arXiv preprint arXiv:2508.03523(2025). https://arxiv.org/abs/2508.03523

2025 arXiv
[16]

https://arxiv.org/abs/2003.11080

Hu, J., Ruder, S., Siddhant, A., Neubig, G., Firat, O., Johnson, M.: XTREME: A Massively Multilingual Multi-task Benchmark for Evaluating Cross-lingual General- ization.arXiv preprint arXiv:2003.11080(2020). https://arxiv.org/abs/2003.11080

2003 arXiv
[17]

https://openai.com/ index/introducing-gpt-5/

OpenAI: Introducing GPT-5.OpenAI Blog, August 7, 2025. https://openai.com/ index/introducing-gpt-5/

2025

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [1]

https: //www.nature.com/articles/s41598-025-98483-1

Raza, M., Jahangir, Z., Riaz, M.B., Saeed, M.J., Sattar, M.A.: Industrial ap- plications of large language models.Scientific Reports15, 98483 (2025). https: //www.nature.com/articles/s41598-025-98483-1

work page 2025

[2] [2]

https: //www.nature.com/articles/s41586-024-07421-0

Farquhar, S., Kossen, J., Kuhn, L., Gal, Y.: Detecting hallucinations in large lan- guage models using semantic entropy.Nature630(8017), 625–630 (2024). https: //www.nature.com/articles/s41586-024-07421-0

work page 2024

[3] [3]

https://arxiv.org/abs/2311.05232

Huang, L., Yu, W., Ma, W., Zhong, W., Feng, Z., Wang, H., Chen, Q., Peng, W., Feng, X., Qin, B., Liu, T.: A survey on hallucination in large language models: prin- ciples, taxonomy, challenges, and open questions.arXiv preprint arXiv:2311.05232 (2024). https://arxiv.org/abs/2311.05232

arXiv 2024

[4] [4]

https://arxiv.org/abs/2109

Lin, S., Hilton, J., Evans, O.: TruthfulQA: Measuring how models mimic human falsehoods.arXiv preprint arXiv:2109.07958(2022). https://arxiv.org/abs/2109. 07958

arXiv 2022

[5] [5]

LessWrong(2025)

Evans, O., Chua, J., Lin, S.: New, improved multiple-choice TruthfulQA. LessWrong(2025). https://www.lesswrong.com/posts/Bunfwz6JsNd44kgLT/ new-improved-multiple-choice-truthfulqa 14 Nery et al

work page 2025

[6] [6]

Association for Computational Lin- guistics

¨Ust¨ un, A., Aryabumi, V., Yong, Z., Ko, W., D’souza, D., Onilude, G., Bhandari, N., Singh, S., Ooi, H.-L., Kayid, A., Vargus, F., Blunsom, P., Longpre, S., Muennighoff, N., Fadaee, M., Kreutzer, J., Hooker, S.: Aya Model: An Instruction Finetuned Open-Access Multilingual Language Model.Proceedings of the 62nd Annual Meet- ing of the Association for Comp...

work page 2024

[7] [7]

https://commoncrawl.github.io/ cc-crawl-statistics/plots/languages

Common Crawl: Statistics of Common Crawl Monthly Archives – Lan- guages.Common Crawl GitHub Pages(2025). https://commoncrawl.github.io/ cc-crawl-statistics/plots/languages

work page 2025

[8] [8]

https://arxiv.org/abs/2310.16787

Longpre, S., Mahari, R., Chen, A., Obeng-Marnu, N., Sileo, D., Brannon, W., Muennighoff, N., Khazam, N., Kabbara, J., Perisetla, K., Wu, X., Shippole, E., Bollacker, K., Wu, T., Villa, L., Pentland, S., Hooker, S.: The Data Provenance Initiative: A Large Scale Audit of Dataset Licensing & Attribution in AI.arXiv preprint arXiv:2310.16787(2023). https://ar...

arXiv 2023