Pith. sign in

REVIEW 4 major objections 5 minor 25 references

Seed-Free Synthetic Data Generation Framework for Instruction-Tuning LLMs: A Case Study in Thai

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A seed-free synthetic dataset of 5,000 Thai instructions, mixing fluency, diversity, and cultural context, matches or beats Thai LLMs trained on 10-100x more data on WangchanThaiInstruct.

desk verdict A solid seed-free synthetic data pipeline for Thai instruction-tuning, but the parity claim rests on BERTScore and needs significance tests and human eval before it holds. read the letter →

arxiv 2411.15484 v1 pith:OTZWYGK7 submitted 2024-11-23 cs.CL

Parinthapat Pengpun , Can Udomcharoenchaikit , Weerayut Buaphet , Peerat Limkonchotiwat This is my paper · ORCID
classification cs.CL
keywords syntheticdatainstructiontuninglow-resourcelanguagesThaiLLMefficiencyseed-freegenerationfluencydiversityculturalcontextfine-tuning
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 tries to establish that a small, carefully designed synthetic instruction dataset can replace datasets an order of magnitude larger when instruction-tuning an LLM for a low-resource language. It identifies three properties of effective training data — fluency, diversity, and cultural context — and builds a seed-free pipeline that generates topics with an LLM, pulls related contexts from Wikipedia, and writes instruction-response pairs directly in Thai. Tuning the LLaMa-3 8B base model on 5,000 such examples matches or beats Thai models trained on 64,000 to 300,000 examples on the WangchanThaiInstruct benchmark, measured by the BERTScore the benchmark authors recommend. If the claim holds, it means a few thousand well-chosen synthetic examples can substitute for much larger, more expensive datasets, and the recipe could be reused for other languages.

What carries the argument

The load-bearing object is the seed-free generation pipeline and its three controllable levers. For each of 750 randomly generated topics, the pipeline either retrieves a Wikipedia article section or asks the generator LLM to invent a context in a random style, then prompts that LLM to write Thai instruction-response pairs for closed question answering, summarization, conversation, and multiple choice. A diversity-control step embeds all samples with BGE-M3 and removes near-duplicates by cosine similarity threshold. The framework's argument is that fluency, diversity, and cultural context can be turned on and off independently, and the full combination, rather than any single property, is what produces the performance jump.

What would settle it

Have human annotators rate a random sample of WangchanThaiInstruct responses from the 5,000-instruction F+C+D+ model, Typhoon-Instruct, and WangchanX without knowing which model produced them; if human preference clearly ranks the 5k model worst despite its near-equal BERTScore, the competitive-parity claim fails. A lighter, purely computational check is to correlate BERTScore with response length on this benchmark, since the paper itself reports its model generates significantly shorter outputs.

Watch

Extended reading notes

Core claim

The paper's central claim is that its best synthetic dataset, F+C+D+ (fluency plus cultural context plus diversity), achieves competitive performance using only 5,000 instructions, despite being generated with no seed instructions and no machine translation. On the WangchanThaiInstruct Thai Culture test set it reaches BERTScore 69.50, above WangchanX's 68.80 and OpenThaiGPT's 64.50 and below Typhoon-Instruct's 74.10; on the General test set it reaches 73.20, above WangchanX's 72.20 and OpenThaiGPT's 67.50 and below Typhoon's 76.50. The authors read this as confirming that all three properties are necessary, since datasets built with only one property score between 45.90 and 57.30 on the culture set and between 49.50 and 59.70 on the general set, far below the combined dataset.

Load-bearing premise

The load-bearing premise is that BERTScore on the WangchanThaiInstruct benchmark faithfully measures whether a model is competitive; if that metric rewards short, generic answers or misses real quality differences, the paper's headline parity claim would collapse.

Editorial extensions

If this is right

  • A 5,000-example synthetic dataset can deliver WangchanThaiInstruct BERTScore above two of three established Thai LLMs, so the data-reduction factor of 10-100x does not cost the benchmark score.
  • All three key properties are load-bearing: single-property ablations fall far short of the combined dataset, so a minimal recipe for synthetic instruction data should include fluent language, topic diversity, and cultural grounding together.
  • The model trained on 5,000 examples transfers across all seven benchmark tasks (brainstorming, classification, closed QA, creative writing, multiple choice, open QA, summarization), despite the underlying dataset containing only four task templates.
  • The largest remaining gap to Typhoon-Instruct is on open-ended tasks, and the paper ties this to its model's shorter average output length, pointing to multi-turn or longer-form training data as the next lever.

Reading between the lines

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

  • The seed-free property is the part most likely to transfer to other low-resource languages: the pipeline needs a capable multilingual generator and Wikipedia coverage, but not a pre-existing instruction dataset, so it can bootstrap from nothing where human-written seeds do not exist.
  • A natural stress test the paper does not run is a scaling curve — if 5,000 examples already match much larger datasets, it is untested whether 10,000 or 20,000 improve on that or merely plateau.
  • Because the parity claim rests on one benchmark and one metric, a human-preference evaluation is the cheapest decisive check on whether the small model is genuinely competitive outside BERTScore; the paper itself lists human evaluation as missing.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes a seed-free synthetic data generation framework for instruction-tuning LLMs in Thai. The pipeline uses Claude-3 Haiku to generate topics, retrieves related Wikipedia contexts, and creates instruction-response pairs for four tasks (closed QA, summarization, conversation, multiple choice), with a semantic-diversity filtering step. Five 5,000-sample datasets are constructed to combine fluency, diversity, and cultural context in different ways. The authors fine-tune Llama-3 8B with QLoRA on each dataset and evaluate on the WangchanThaiInstruct benchmark. The best dataset (F+C+D+) achieves BERTScore 69.50 on the Thai Culture test set and 73.20 on the General test set, outperforming WangchanX and OpenThaiGPT but trailing Typhoon-Instruct. The central claim is that this 5,000-instruction synthetic dataset performs competitively with state-of-the-art Thai LLMs trained on 10-100x more data.

Significance. If the central claim were established, this would be a practically valuable result for low-resource language instruction tuning: it would show that a small, seed-free, synthetic dataset can substitute for much larger corpora, lowering cost and barrier to entry. The paper is clearly written, the pipeline is described in enough detail to reproduce, and the authors release code and data. The methodological idea of generating diverse topics from scratch and grounding instructions in Wikipedia contexts is reasonable. However, the evidence presented is not yet sufficient to support the abstract's 'competitive performance' claim, because the evaluation depends on a single metric (BERTScore) with small margins and no significance tests against two of three baselines, and the ablation design does not cleanly isolate the three proposed properties.

major comments (4)
  1. [§5.2 Wilcoxon tests] The central claim that F+C+D+ achieves 'competitive performance' is supported only by BERTScore point estimates. On the Thai Culture set the margin over WangchanX is 0.70 points (69.50 vs 68.80) and over OpenThaiGPT is 5.00 points; on the General set the margins are 1.00 and 5.70. No confidence intervals, standard errors, or significance tests are reported for these comparisons. The only Wilcoxon tests in the paper (Section 5.2) compare F+C+D+ against Typhoon, not against WangchanX or OpenThaiGPT. Given the small margins, especially versus WangchanX, the parity claim is not statistically grounded. Moreover, all lexical metrics show large deficits for F+C+D+ (e.g., Thai Culture BLEU 0.10 vs 2.24 for WangchanX and 2.32 for Typhoon; ROUGE-L 7.50 vs 12.60 and 20.00). The paper defers to the benchmark authors' statement that BERTScore is 'the most reliable metric,' but no evidence is provided that BERTScore correlates with human judgment for Thai, nor that it is insensitive to the brevity bias that the paper itself identifies in Section 5.2. To support the abstract claim, the authors should report paired significance tests for all baseline comparisons and, ideally, a human evaluation or a validation of BERTScore against human ratings for Thai.
  2. [§5.2, Figure 2] The Wilcoxon rank-sum tests reported in Section 5.2 compare F+C+D+ against Typhoon-Instruct only, and they find statistically significant differences with an average statistic of -6.512 (p = 0.00073). This documents that the model is significantly worse than the strongest Thai baseline on most tasks, contradicting any interpretation of 'competitive' with Typhoon. Tests against WangchanX and OpenThaiGPT, the models that F+C+D+ is claimed to surpass, are absent. If the margins over those baselines are not statistically significant, the headline result reduces to a non-significant trend. The authors should report paired significance tests for all three baseline comparisons, with appropriate multiple-comparison corrections.
  3. [§4.1] The ablations designed to show that all three properties are necessary are confounded. F+C+D+ uses 750 topics (400 cultural, 300 general), while F+C-D- ('Fluency Only') uses only 10 general topics and no diversity control. This varies topic count and diversity simultaneously, so the performance gap between these two datasets cannot be attributed specifically to cultural context or diversity. Similarly, the 'Cultural Context Only' dataset is built by taking 1,000 samples from F+C+D+, translating to English, paraphrasing four times, and translating back to Thai; this changes diversity, fluency, and cultural grounding at the same time. The 'Diversity Only' dataset uses back-translation, which affects fluency and may also remove Thai cultural references. The claims that 'all three properties are required' (Section 5.1) and the corresponding abstract-level conclusions are not supported by these manipulations. The authors should design ablations that vary one property while holding others fixed, for example by generating F+C+D- and F+C-D+ with the same 750-topic protocol and only changing the culture/topic mix.
  4. [Limitations] The Limitations section explicitly acknowledges that 'conducting extensive human evaluations would be crucial for assessing the practical usability and perceived quality of the generative models.' Given that the paper's central claim is about competitive performance, and the objective metric used (BERTScore) is disputed in terms of validity for Thai, the absence of human evaluation is not a peripheral issue but a load-bearing gap. The paper should either provide human judgments or a strong external validation of BERTScore for Thai instruction following, or substantially soften the abstract claim.
minor comments (5)
  1. [§1, Abstract] The phrase 'state-of-the-art Thai LLMs trained on hundreds of thousands of instructions' is imprecise: WangchanX uses 64k examples, and OpenThaiGPT's current dataset size is not reported, while Typhoon's instruction-tuning data size is not disclosed. Please state the actual sizes for each baseline.
  2. [§3.3] The cultural topics prompt is truncated with 'The rest of this prompt is omitted as it is the same as the General Topics Prompt.' For reproducibility, the full prompt should appear in an appendix.
  3. [§5.2] The Wilcoxon statistic is reported as a negative average (-6.512), which is unconventional because W is usually non-negative. Please clarify whether these are standardized test statistics or the sum of ranks with a sign convention, and report effect sizes.
  4. [Table 2] The table formatting is hard to read in the provided version: the header spans two lines and the bold/underline highlighting is not visible. A cleaner table with clear markers for the best and second-best values would improve readability.
  5. [§6] The conclusion says the model 'beats' OpenThaiGPT and is 'comparable' to WangchanX, but the abstract uses 'competitive performance' against state-of-the-art LLMs. Please reconcile these wordings to avoid overclaiming in the abstract.

Circularity Check

1 steps flagged · score 4.0 of 10

No constructional circularity in the data pipeline; the only circular-adjacent element is a load-bearing self-benchmark metric choice that makes the 5k model look competitive.

  1. self citation load bearing [Section 4.3 (Metrics); relied upon in Section 5.1 Discussion and the Abstract]
    "However, we do note that the WangchanX-10k mentioned that BERTScore is the most reliable metric as it measures semantic similarity, while other traditional metrics yield inconclusive results."

    The abstract's central claim (competitive performance using only 5,000 instructions) is operationalized in Section 5.1 solely as the second-highest BERTScore on WangchanThaiInstruct, while every other reported metric shows large deficits for F+C+D+ (e.g., BLEU 0.10 vs 2.32, METEOR 6.70 vs 11.30, ROUGE-L 7.50 vs 20.00 against Typhoon on the Thai Culture set). The paper justifies this exclusive reliance on BERTScore by deferring to 'WangchanX-10k', an evaluation suite associated with the same research group as two of the present authors (Phatthiyaphaibun et al. 2024 lists Limkonchotiwat and Udomcharoenchaikit).

full rationale

The derivation chain is mostly self-contained. The synthetic instruction data are generated from randomly sampled topics plus Wikipedia contexts, with no use of the evaluation benchmark's training split, and no fitted parameter is tuned to the evaluation outcome, so the main data-efficiency result is not circular by construction. The three key properties (fluency, diversity, cultural context) are introduced as a hypothesis and tested through five ablations; the ablations are confounded because topic count, back-translation, and base-sample count vary jointly, but confounding is an experimental validity issue rather than circularity. The one load-bearing circular-adjacent element is the metric choice: the abstract's competitive-performance claim rests on BERTScore, and the paper justifies treating BERTScore as 'the most reliable metric' by appealing to WangchanX-10k, an evaluation suite from the same research group as two of the present authors. Without that self-imported BERTScore preference, the central claim is contradicted by the lexical metrics reported in Table 2. The Limitations section honestly concedes that no human evaluation was conducted, which reinforces the concern. However, because the benchmark's reference answers are human-annotated and external, and the finetuning experiment itself is new work, the paper retains independent content beyond the self-citation, so the circularity score is moderate rather than high.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claim depends less on numerical fitting than on measurement assumptions and dataset-construction assumptions. The free parameters are pipeline hyperparameters, not fit to the benchmark. The most fragile items are the domain assumptions that BERTScore and a single benchmark capture instruction quality, and that the ablations isolate the named properties.

free parameters (4)
  • Diversity deduplication similarity threshold = 0.95
    Hand-chosen cosine threshold used to remove near-duplicate samples; affects dataset size and diversity but is not fitted to the benchmark.
  • Topic counts for F+C+D+ = 750 (400 cultural, 300 general)
    Hand-chosen; the Fluency-only ablation uses 10 topics, so topic count varies across compared datasets.
  • Generation temperatures = 0.95 for topics; 0.35 (QA, summarization), 0.8 (conversation), 0.4 (multiple choice) for instructions
    Hand-chosen sampling temperatures; influence the variety of generated content.
  • Number of paraphrases per sample in C-only and No-Properties datasets = 4
    Hand-chosen to inflate dataset to 5,000 samples; affects diversity manipulation.
assumptions (5)
  • domain assumption BERTScore is a reliable metric for Thai instruction-following quality
    Section 4.3 relies on WangchanX-10k's recommendation that BERTScore is the most reliable among the listed metrics; this is an external claim, not independently verified here.
  • domain assumption WangchanThaiInstruct benchmark validly measures Thai LLM instruction-following
    Section 4.3 uses a single human-annotated benchmark (legal, medical, finance domains) as the sole evaluation of all models.
  • domain assumption Llama-3 8B base model with QLoRA is a fair base for comparing against models that may use continual pretraining
    The paper fine-tunes only the base Llama-3 8B without continual pretraining, while baselines like Typhoon use continual pretraining; comparability is assumed.
  • domain assumption The three properties (fluency, diversity, cultural context) are the key determinants of instruction dataset quality
    Section 1 motivates this from examples in existing Thai datasets; it is the hypothesis being tested, not an established fact.
  • ad hoc to paper Back-translation and paraphrasing selectively remove only the intended property
    Section 4.1 assumes NLLB round-trip reduces fluency and QCPG paraphrasing reduces diversity, but these operations also alter content and length.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Seed-Free Synthetic Data Generation Framework for Instruction-Tuning LLMs: A Case Study in Thai." pith.science (2026). https://pith.science/paper/OTZWYGK7

@misc{pith2026241115484,
  author       = {Pith},
  title        = {Pith review of: Seed-Free Synthetic Data Generation Framework for Instruction-Tuning LLMs: A Case Study in Thai},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OTZWYGK7}},
  note         = {Machine review of arXiv:2411.15484}
}
read the original abstract

We present a synthetic data approach for instruction-tuning large language models (LLMs) for low-resource languages in a data-efficient manner, specifically focusing on Thai. We identify three key properties that contribute to the effectiveness of instruction-tuning datasets: fluency, diversity, and cultural context. We propose a seed-data-free framework for generating synthetic instruction-tuning data that incorporates these essential properties. Our framework employs an LLM to generate diverse topics, retrieve relevant contexts from Wikipedia, and create instructions for various tasks, such as question answering, summarization, and conversation. The experimental results show that our best-performing synthetic dataset, which incorporates all three key properties, achieves competitive performance using only 5,000 instructions when compared to state-of-the-art Thai LLMs trained on hundreds of thousands of instructions. Our code and dataset are publicly available at https://github.com/parinzee/seed-free-synthetic-instruct.

Figures

Figures reproduced from arXiv: 2411.15484 by the authors.

Figure 1
Figure 1. Our proposed framework for generating syn [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 3
Figure 3. Comparison of average generation lengths [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

25 extracted references · 8 canonical work pages

  1. [4]

    Preprint, arXiv:2404.05829

    Sam- balingo: Teaching large language models new lan- guages. Preprint, arXiv:2404.05829. Ning Ding, Yulin Chen, Bokai Xu, Yujia Qin, Shengding Hu, Zhiyuan Liu, Maosong Sun, and Bowen Zhou

    arXiv

  2. [5]

    In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 3029–3051, Singapore

    Enhancing chat language mod- els by scaling high-quality instructional conversa- tions. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 3029–3051, Singapore. Association for Com- putational Linguistics. Qianlong Du, Chengqing Zong, and Jiajun Zhang

    work page 2023

  3. [6]

    Preprint, arXiv:2311.15653

    Mods: Model-oriented data selection for instruction tuning. Preprint, arXiv:2311.15653. Maxim Enis and Mark Hopkins

    arXiv

  4. [7]

    Preprint, arXiv:2404.13813

    From llm to nmt: Advancing low-resource machine translation with claude. Preprint, arXiv:2404.13813. Zhiqiang Hu, Nancy Chen, and Roy Lee

    arXiv

  5. [8]

    In Findings of the As- sociation for Computational Linguistics: EMNLP 2023, pages 693–703, Singapore

    Adapter- TST: A parameter efficient method for multiple- attribute text style transfer. In Findings of the As- sociation for Computational Linguistics: EMNLP 2023, pages 693–703, Singapore. Association for Computational Linguistics. Haoyang Huang, Tianyi Tang, Dongdong Zhang, Xin Zhao, Ting Song, Yan Xia, and Furu Wei

    work page 2023

  6. [9]

    In Findings of the Association for Com- putational Linguistics: EMNLP 2023, pages 12365– 12394, Singapore

    Not all languages are created equal in LLMs: Improv- ing multilingual capability by cross-lingual-thought prompting. In Findings of the Association for Com- putational Linguistics: EMNLP 2023, pages 12365– 12394, Singapore. Association for Computational Linguistics. Lea Krause, Wondimagegnhue Tufa, Selene Baez San- tamaria, Angel Daza, Urja Khurana, and P...

    work page 2023

  7. [10]

    Preprint, arXiv:2309.14517

    Watch your language: Investigat- ing content moderation with large language models. Preprint, arXiv:2309.14517. Shayne Longpre, Gregory Yauney, Emily Reif, Kather- ine Lee, Adam Roberts, Barret Zoph, Denny Zhou, Jason Wei, Kevin Robinson, David Mimno, and Daphne Ippolito

    arXiv

  8. [11]

    Preprint, arXiv:2305.13169

    A pretrainer’s guide to training data: Measuring the effects of data age, domain coverage, quality, & toxicity. Preprint, arXiv:2305.13169. Yecheng Jason Ma, William Liang, Guanzhi Wang, De- An Huang, Osbert Bastani, Dinesh Jayaraman, Yuke Zhu, Linxi Fan, and Anima Anandkumar

    arXiv

Show all 25 references
  1. [12]

    Preprint, arXiv:2310.12931

    Eu- reka: Human-level reward design via coding large language models. Preprint, arXiv:2310.12931. Xuan-Phi Nguyen, Wenxuan Zhang, Xin Li, Mahani Aljunied, Qingyu Tan, Liying Cheng, Guanzheng Chen, Yue Deng, Sen Yang, Chaoqun Liu, Hang Zhang, and Lidong Bing

    arXiv

  2. [13]

    Preprint, arXiv:2312.00738

    Seallms – large language models for southeast asia. Preprint, arXiv:2312.00738. OpenAI, Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Ale- man, Diogo Almeida, Janko Altenschmidt, Sam Alt- man, Shyamal Anadkat, Red Avila, Igor Babuschkin,...

    arXiv

  3. [14]

    Preprint, arXiv:2303.08774

    Gpt-4 technical report. Preprint, arXiv:2303.08774. OpenThaiGPT

    arXiv

  4. [15]

    Preprint, arXiv:2403.16127

    Wangchanlion and wangchanx mrc eval. Preprint, arXiv:2403.16127. Kunat Pipatanakul, Phatrasek Jirabovonvisut, Potsawee Manakul, Sittipong Sripaisarnmongkol, Ruangsak Patomwong, Pathomporn Chokchainant, and Kasima Tharnpipitchai

    arXiv

  5. [16]

    Preprint, arXiv:2312.13951

    Typhoon: Thai large language models. Preprint, arXiv:2312.13951. Alec Radford, Jeff Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever

    arXiv

  6. [17]

    In Proceedings of the First Workshop on Patient-Oriented Language Pro- cessing (CL4Health) @ LREC-COLING 2024, pages 124–130, Torino, Italia

    Building certified medical chatbots: Overcoming unstructured data lim- itations with modular RAG. In Proceedings of the First Workshop on Patient-Oriented Language Pro- cessing (CL4Health) @ LREC-COLING 2024, pages 124–130, Torino, Italia. ELRA and ICCL. AI Singapore

    2024

  7. [18]

    In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 1941–1961, Singapore

    CLASS: A design frame- work for building intelligent tutoring systems based on learning science principles. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 1941–1961, Singapore. Association for Computational Linguistics. Gemini Team, Rohan Anil,...

    2023

  8. [19]

    Preprint, arXiv:2312.11805

    Gemini: A fam- ily of highly capable multimodal models. Preprint, arXiv:2312.11805. NLLB Team, Marta R. Costa-jussà, James Cross, Onur Çelebi, Maha Elbayad, Kenneth Heafield, Kevin Hef- fernan, Elahe Kalbassi, Janice Lam, Daniel Licht, Jean Maillard, Anna Sun, Skyler Wang, Gui...

    arXiv

  9. [21]

    Preprint, arXiv:2212.10560

    Self-instruct: Aligning language models with self-generated instructions. Preprint, arXiv:2212.10560. Frank Wilcoxon

    arXiv

  10. [22]

    Preprint, arXiv:2304.12244

    Wizardlm: Empowering large language models to follow complex instructions. Preprint, arXiv:2304.12244. Boyang Xue, Hongru Wang, Weichao Wang, Rui Wang, Sheng Wang, Zeming Liu, and Kam-Fai Wong

    arXiv

  11. [23]

    Preprint, arXiv:2402.13606

    A comprehensive study of multilingual confidence estimation on large language models. Preprint, arXiv:2402.13606. Xiang Zhang, Senyu Li, Bradley Hauer, Ning Shi, and Grzegorz Kondrak

    arXiv

  12. [24]

    In Proceedings of the 2023 Conference on Empirical Methods in Natu- ral Language Processing, pages 7915–7927, Singa- pore

    Don’t trust ChatGPT when your question is not in English: A study of multilin- gual abilities and types of LLMs. In Proceedings of the 2023 Conference on Empirical Methods in Natu- ral Language Processing, pages 7915–7927, Singa- pore. Association for Computational Linguistics...

    2023

  13. [25]

    Preprint, arXiv:2305.11206

    Lima: Less is more for alignment. Preprint, arXiv:2305.11206. A Prompts for Each Task in Instruction Generation Closed Question Answering: Generate 5 questions focusing on differ- ent aspects / parts of this given context. Use only the given context to create your questions. D...

    arXiv

  14. [2019]

    Development and deployment of a large-scale dialog-based intelligent tutoring sys- tem. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Com- putational Linguistics: Human Language Technolo- gies, Volume 2 (Industry Papers) , pages 114...

    2019

  15. [2022]

    Preprint, arXiv:2207.04672

    No language left behind: Scal- ing human-centered machine translation. Preprint, arXiv:2207.04672. Kobkrit Viriyayudhakorn and Charin Polpanumas

    arXiv

  16. [2023]

    In Proceedings of the 2023 Conference on Empir- ical Methods in Natural Language Processing, pages 4232–4267, Singapore

    MEGA: Multilingual evaluation of generative AI. In Proceedings of the 2023 Conference on Empir- ical Methods in Natural Language Processing, pages 4232–4267, Singapore. Association for Computa- tional Linguistics. AI@Meta

    2023

  17. [2024]

    Preprint, arXiv:2402.03216

    Bge m3-embedding: Multi-lingual, multi-functionality, multi-granularity text embeddings through self-knowledge distillation. Preprint, arXiv:2402.03216. Mike Conover, Matt Hayes, Ankit Mathur, Jianwei Xie, Jun Wan, Sam Shah, Ali Ghodsi, Patrick Wendell, Matei Zaharia, and Reynold Xin

    arXiv

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.