Pith. sign in

REVIEW 2 major objections 6 minor 32 references

Teaching a Language Model to Speak the Language of Tools

T0 review · 2 major / 6 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 Fine-tuning on a bilingual dataset of 10,035 conversations lifts Bulgarian models' tool-calling accuracy by up to 28.75 percentage points without harming their language benchmarks.

desk verdict Useful Bulgarian function-calling artifacts, but the headline accuracy numbers rest on a 120-case self-built eval with no leakage check; the core idea likely holds. read the letter →

arxiv 2506.23394 v1 pith:P5YOWLXJ submitted 2025-06-29 cs.IR cs.AIcs.CL

Simeon Emanuilov This is my paper
classification cs.IRcs.AIcs.CL
keywords functioncallingtooluseLoRAfine-tuningmultilinguallanguagemodelsBulgarianlow-resourceadaptationModelContextProtocolbenchmarkevaluation
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 tries to establish that a language model can be taught to call external tools reliably in a non-English language by continued fine-tuning on a small bilingual dataset, without erasing its existing language skills. Using Bulgarian as the case study, the author adapts the BgGPT model family at three scales, 2.6B, 9B, and 27B parameters, with Low-Rank Adaptation (LoRA) on 10,035 conversations in which tool schemas are in English while user and assistant turns are in Bulgarian. On a 120-case evaluation spanning six tool-use scenarios, the fine-tuned TUCAN models beat their base models by 28.75, 8.34, and 0.83 percentage points, with the smallest model improving most. The broader point is a recipe: if this holds, the same pipeline can extend tool-augmented AI to languages that today's English-centric function-calling ecosystem largely ignores.

What carries the argument

The load-bearing mechanism is a parameter-efficient continued-training recipe: Low-Rank Adaptation (LoRA) fine-tunes only 0.79% to 1.2% of each BgGPT model's parameters for three epochs. The dataset is the other half: 10,035 bilingual conversations that pair English function definitions with Bulgarian dialogue, mark tools, calls, and results with a lightweight tag convention, and deliberately include negative cases where the correct behavior is to decline a tool or request clarification. The same structured prompt template is enforced at training and inference, and the released Tucan-Eval harness grades outputs into five error classes (no call when expected, unexpected call, wrong function, wrong parameters, malformed JSON). Together these pieces convert 'whether and when to call' from a prompted instruction into a learned behavior.

What would settle it

Write a new Bulgarian function-calling test set independently of the 10,035 training conversations — fresh functions, fresh queries, no topic or phrasing overlap — and rerun the six scenario types on TUCAN-2.6B and its base model. If the 28.75-point gap collapses toward the base model's level on this held-out set, the improvement is format memorization; if it survives, fine-tuning installed a generalizable skill. A cheaper first step is a near-duplicate scan (n-gram or embedding similarity) between the 120 evaluation cases and the training corpus.

Watch

Extended reading notes

Core claim

The paper's central claim is that specialized fine-tuning, not prompt engineering, is what gives a language model dependable function-calling behavior in Bulgarian. TUCAN-2.6B reaches 78.75% overall accuracy versus 50.00% for its base model; TUCAN-9B reaches 86.67% versus 78.33%; and TUCAN-27B reaches 87.50% versus 86.67%, averaged over multiple runs of the six-scenario benchmark. The author attributes the gains to training the model on when to call a tool, when to decline one, and when to ask for missing parameters, using a fixed prompt template with XML-style tags (<tools>, <tool_code>, <tool_response>). Knowledge-retention checks on HellaSwagBG, WinograndeBG, ARC-Easy-BG, and ARC-Challenge-BG show deviations within roughly ±0.04 points, which the paper reads as measurement noise rather than catastrophic forgetting. The paper further claims a production benefit: TUCAN emits clean, parseable JSON function calls, whereas base models surround their calls with verbose prose that complicates machine consumption.

Load-bearing premise

The 120 test cases in the Bulgarian evaluation set are an unbiased measure of function-calling skill — specifically, that they do not overlap with or share a tell-tale format with the 10,035 training conversations, so the reported accuracy gains are genuine tool-use competence rather than memorization of the training template.

Editorial extensions

If this is right

  • The full pipeline — dataset, models, LoRA adapters, and evaluation framework — is released openly, so another language needs only its own bilingual function-calling corpus to repeat the experiment.
  • Smaller models gain the most (28.75 points for 2.6B versus 0.83 points for 27B), so the recipe is cheapest exactly where base tool-use ability is weakest.
  • TUCAN outputs are directly machine-parseable, removing the post-processing step that verbose base-model responses require in production agents.
  • Residual failures concentrate in wrong or incomplete parameters rather than malformed JSON, identifying parameter extraction as the next bottleneck for training data.
  • Because the training format follows MCP-style structured tool definitions, the adapted models integrate with standardized agent protocols rather than proprietary interfaces.

Reading between the lines

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

  • The sharp inverse scaling of the gains suggests the small model was missing a discrete behavioral rule — call versus do not call — that fine-tuning installs; a natural extension is to test whether even smaller models or other language families show the same jump.
  • The cleanest external check of the claim is a fresh Bulgarian evaluation set written independently of the 10,035 training conversations; if the large gains persist on such a distribution-shifted set, the competence is general rather than template-specific.
  • The paper keeps tool schemas in English while localizing dialogue; languages with their own developer ecosystems would likely need a second pass that localizes function names and parameter descriptions, which the released dataset format already supports.
  • The single-model-family comparison leaves open how much of the gain is specific to fine-tuning versus formatting; benchmarking prompt-engineered versions of the same base models would separate those effects.
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

2 major / 6 minor

Summary. The paper presents TUCAN, a set of LoRA fine-tuned BgGPT-Gemma-2 models (2.6B, 9B, and 27B) trained on a newly constructed bilingual Bulgarian/English function-calling dataset of 10,035 conversations. The authors report accuracy improvements on their own 120-case Tucan-BG-Eval-v1.0 benchmark, up to +28.75 percentage points for the 2.6B model, and claim that core language understanding is preserved on Bulgarian HellaSwag, Winogrande, and ARC benchmarks. The models, dataset, and evaluation framework are released for replication.

Significance. If the reported results are secure, this is a practically useful contribution: it provides a replicable recipe for adding function-calling behavior to a non-English language model family, with released artifacts and a direct base-versus-fine-tuned comparison. The knowledge-retention check on established Bulgarian benchmarks, with standard errors reported, is a genuine strength, as is the open release of the training data and evaluation harness. However, the central quantitative claim depends on a small, self-authored evaluation set with no demonstrated separation from the training data, and the retention analysis in Section 6.4 contains arithmetic inconsistencies. The contribution's significance is therefore conditional on correcting these issues.

major comments (2)
  1. [§5.1.5 and §6.1] The headline function-calling gains rest entirely on the Tucan-BG-Eval-v1.0 dataset of 120 cases, but the paper does not report any overlap or leakage analysis between these cases and the 10,035 training conversations. No n-gram overlap, template-distance, or schema-overlap check is described, and both sets use the same XML/tool_call format. With 120 cases, one case equals 0.83 percentage points, so the 27B improvement is exactly one test case and the 2.6B improvement corresponds to about 34.5 cases. Under these conditions, the reported improvements could reflect format memorization rather than generalizable tool-use competence. Please provide a quantitative leakage analysis, report confidence intervals or run-level variance for the accuracy scores, and ideally validate on an independently constructed or held-out set.
  2. [§6.4 and Table 6] The narrative in Section 6.4 is inconsistent with the numbers in Table 6. The text claims a maximum deviation of 0.0382 on HellaSwagBG for the 2.6B model, but Table 6 gives a difference of only 0.0053 for that cell. It also claims a WinograndeBG gain of +0.0635 for the 2.6B model, while the table shows +0.0150. The stated average improvement of +0.0176 for the 2.6B model does not match the table values, which average +0.0059. These discrepancies directly affect the 'no catastrophic forgetting' conclusion and must be corrected, with all derived statements recomputed accordingly.
minor comments (6)
  1. [§3 and §4] Section 3.6 appears twice, once for topic distribution and once for message length, and Section 4.3 appears twice, once for the prompt template and once for hyperparameters; please renumber these sections.
  2. [§6.1] Table 3 is used both for the hyperparameter configuration and for the function-calling accuracy results; the accuracy table should be renumbered to avoid ambiguity.
  3. [§6.1 and §6.2] The in-text figure references appear to be off by one: Section 6.1 refers to 'Figure 5' for the overall accuracy results, while the corresponding caption is Figure 4, and Section 6.2 refers to 'Figure 6' for scenario breakdown, while the corresponding caption is Figure 5.
  4. [§4.3] In the prompt template, the token `toll_response` appears to be a typo for `tool_response`; please verify whether this is a typo in the manuscript or in the actual template used for training.
  5. [§1 and §2.1] The benchmark name 'Winogrande' is misspelled as 'Winograde' in the Introduction and Section 2.1.
  6. [§6.7] The discussion already acknowledges that the 120-case evaluation represents controlled conditions; this limitation statement is appropriate and should be retained after the leakage analysis is added.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the reported gains are direct base-vs-TUCAN comparisons on a fixed evaluation set, not outputs of the training procedure by construction.

full rationale

The paper's central claim is that LoRA fine-tuning on Tucan-BG-v1.0 improves function-calling accuracy relative to the corresponding BgGPT base models, as measured by Tucan-BG-Eval-v1.0, while preserving performance on external Bulgarian benchmarks. No step in this chain is circular by construction: the training data (10,035 conversations) and the evaluation set (120 curated cases) are separate artifacts, the accuracy numbers are obtained by running both base and fine-tuned models through the same parser and comparing against fixed expected behaviors, and the knowledge-retention checks use the externally maintained lm-evaluation-harness-bg benchmarks rather than the author's own data. The main validity concern is that the evaluation set was authored by the same pipeline that produced the training data and shares the same <tool_call> format, and the paper reports no overlap or leakage analysis (Sections 3.1, 4.3, 5.1.2, 5.1.5). That is a benchmark-independence threat, not a circular reduction: the evaluation does not use training labels as predictions, and no fitted parameter is renamed as a result. The paper itself acknowledges the laboratory nature of the 120-case set (Section 6.7). Separately, the knowledge-retention prose in Section 6.4 is arithmetically inconsistent with Table 6 (e.g., it claims a WinograndeBG gain of +0.0635 and an average of +0.0176 for Tucan-2.6B, whereas the table shows +0.0150 and +0.0059), which undermines the 'no forgetting' claim as stated but is an internal-consistency error, not circularity. Weighing these, there is no exhibited derivation step that reduces to its own input.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central accuracy claim rests on manually chosen hyperparameters and on unverified assumptions about data quality, evaluation validity, and train/eval separation. The released artifacts allow others to check some of these assumptions, but the paper itself does not.

free parameters (6)
  • LoRA rank (r) = 16 (2.6B), 32 (9B, 27B)
    Chosen by hand after the 2.6B model showed instability at higher ranks; directly controls adaptation capacity and influences all reported accuracy gains (Section 4.3).
  • LoRA alpha = 16 (2.6B), 32 (9B, 27B)
    Set equal to rank for each model; a manual choice that affects the scale of updates and thus the fine-tuning outcome (Section 4.3).
  • Training epochs = 3
    Fixed across all model sizes with no ablation; the central results depend on this unexamined choice (Table 3 in Section 4.3).
  • Synthetic data generation model mixture = GPT-4.1 40%, Gemini 2.5 Pro 30%, Claude Sonnet 4 30%
    A manually chosen composition of generator models; it shapes the style and content of the 10,035 training conversations (Section 3.1).
  • Parameter-matching leniency in evaluation = lenient type matching, string-to-number coercion, normalization
    The Tucan-Eval comparison rules affect whether arguments count as correct; no sensitivity analysis is provided (Section 5.1.2).
  • Evaluation test set size and scenario allocation = 120 cases, 20 per scenario
    Chosen by hand; with 20 cases per scenario, per-scenario percentages move in 5% steps and the 27B overall gain is one case (Section 5.1.5).
assumptions (4)
  • domain assumption The 10,035 training conversations are high quality and representative of real-world Bulgarian tool use.
    Dataset is a mix of manual gold examples and proprietary LLM outputs; no human quality audit, inter-annotator agreement, or independent validation is reported (Section 3.1).
  • domain assumption The 120-case Tucan-BG-Eval-v1.0 set is a valid and unbiased measure of function-calling capability.
    The eval set is authored by the same pipeline as the training data and no leakage check is described (Section 5.1.5).
  • domain assumption The translated Bulgarian benchmarks (HellaSwagBG, WinograndeBG, ARC) measure the same capabilities as their English originals.
    Knowledge retention is assessed with these benchmarks, but their translation validity is not established in the paper (Section 5.2.1).
  • domain assumption Base BgGPT models are representative state-of-the-art Bulgarian models for this comparison.
    The comparison is only against BgGPT base models; no other multilingual models or prompt baselines are included (Section 6.7).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Teaching a Language Model to Speak the Language of Tools." pith.science (2026). https://pith.science/paper/P5YOWLXJ

@misc{pith2026250623394,
  author       = {Pith},
  title        = {Pith review of: Teaching a Language Model to Speak the Language of Tools},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/P5YOWLXJ}},
  note         = {Machine review of arXiv:2506.23394}
}
read the original abstract

External tool integration through function-calling is essential for practical language model applications, yet most multilingual models lack reliable tool-use capabilities in non-English languages. Even state-of-the-art multilingual models struggle with determining when to use tools and generating the structured outputs required for function calls, often exhibiting language confusion when prompted in lower-resource languages. This work presents a methodology for adapting existing language models to enable robust tool use in any target language, using Bulgarian as a case study. The approach involves continued training of the BgGPT model series (2.6B, 9B, 27B parameters) on a novel bilingual dataset of 10,035 function-calling examples designed to support standardized protocols like MCP (Model Context Protocol). The research introduces TUCAN (Tool-Using Capable Assistant Navigator), which achieves up to 28.75% improvement in function-calling accuracy over base models while preserving core language understanding, as verified on established Bulgarian benchmarks. Beyond accuracy gains, TUCAN models demonstrate production-ready response formatting with clean, parsable function calls, contrasting with the verbose and inconsistent outputs of base models. The models, evaluation framework, and dataset are released to enable replication for other languages. This work demonstrates a practical approach for extending tool-augmented capabilities beyond English-centric systems.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

32 extracted references · 11 canonical work pages

  1. [1]

    Introduction The emergence of language models has fundamentally transformed how we interact with artificial intelligence, with open-source developments rapidly approaching the capabilities of proprietary systems (Zhang, et al. 2024). However, the true transformative potential of these models extends far beyond their inherent linguistic competencies—it lie...

    work page 2024

  2. [3]

    2025), Meta's Llama series (Grattafiori, et al

    (Team, Kamath, et al. 2025), Meta's Llama series (Grattafiori, et al. 2024), and Alibaba's Qwen models (Yang, et al

    work page 2025

  3. [5]

    Branch-and-Merge

    extend language coverage to 500+ languages through continual pre-training, while studies show that targeted multilingual adaptation can outperform both monolingual and massively multilingual baselines for related language families (Downey, et al. 2024). The Bulgarian language modeling shows successful language-specific adaptation through the BgGPT series ...

    work page 2024

  4. [6]

    However, both frameworks remain predominantly English-centric

    introduces stateful, conversational evaluation scenarios that better reflect real-world deployment conditions. However, both frameworks remain predominantly English-centric. Contemporary research addresses critical challenges, including determining when tools are necessary versus relying on internal knowledge (MetaTool benchmark) (Huang, et al. 2023), han...

    work page 2023

  5. [8]

    First, a specialized framework was developed to measure function-calling accuracy in multilingual contexts

    Evaluations To assess the effectiveness of the fine-tuning methodology, a two-part evaluation was conducted. First, a specialized framework was developed to measure function-calling accuracy in multilingual contexts. Second, knowledge retention was evaluated on standard Bulgarian language benchmarks to verify that enhanced tool-use capabilities did not co...

    work page 2020

  6. [10]

    arXiv preprint arXiv:2410.14815

    Adapting Multilingual LLMs to Low-Resource Languages using Continued Pre-training and Synthetic Corpus. arXiv preprint arXiv:2410.14815. Rafailov, Rafael, Archit Sharma, Eric Mitchell, Christopher D. Manning, Stefano Ermon, and Chelsea Finn

    arXiv

  7. [11]

    arXiv preprint arXiv:2205.12393

    Fine-tuned language models are continual learners. arXiv preprint arXiv:2205.12393. Wolf, Thomas, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, and Pierric Cistac

    arXiv

  8. [14]

    arXiv preprint arXiv:2401.01055

    LLaMA Beyond English: An Empirical Study on Language Capability Transfer. arXiv preprint arXiv:2401.01055. Sakaguchi, Keisuke, Ronan Le Bras, Chandra Bhagavatula, and Yejin Choi

    arXiv

Show all 32 references
  1. [16]

    arXiv:2503.23278

    Model Context Protocol (MCP): Landscape, Security Threats, and Future Research Directions . arXiv:2503.23278. Zhang, Gongbo, Qiao Jin, Yiliang Zhou, Song Wang, Betina Idnay, Yiming Luo, and Elizabeth Park

    arXiv

  2. [17]

    arXiv preprint arXiv:2501.10868

    JSONSchemaBench: A Rigorous Benchmark of Structured Outputs for Language Models. arXiv preprint arXiv:2501.10868. Team, Gemma, Morgane Riviere, Shreya Pathak, Pier Giuseppe Sessa, Cassidy Hardin, Surya Bhupatiraju, and Léonard Hussenot

    arXiv

  3. [18]

    arXiv preprint arXiv:2408.00118

    Gemma 2: Improving Open Language Models at a Practical Size . arXiv preprint arXiv:2408.00118. Team, Gemma, Aishwarya Kamath, Johan Ferret, Shreya Pathak, Nino Vieillard, Ramona Merhej, and Sarah Perrin

    arXiv

  4. [19]

    arXiv preprint arXiv:2503.19786

    Gemma 3 Technical Report. arXiv preprint arXiv:2503.19786. Grattafiori, Aaron, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, and Aiesha Letman

    arXiv

  5. [20]

    arXiv preprint arXiv:2407.21783

    The Llama 3 Herd of Models. arXiv preprint arXiv:2407.21783. Yang, An, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, and Bowen Yu

    arXiv

  6. [21]

    arXiv preprint arXiv:2505.09388

    Qwen3 Technical Report. arXiv preprint arXiv:2505.09388 . Conneau, Alexis, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettlemoyer, and Veselin Stoyanov

    arXiv

  7. [22]

    arXiv preprint arXiv:1911.02116

    Unsupervised Cross-lingual Representation Learning at Scale . arXiv preprint arXiv:1911.02116. Ji, Shaoxiong, Zihao Li, Indraneil Paul, Jaakko Paavola, Peiqin Lin, Pinzhen Chen, and Dayyán O'Brien

    1911 arXiv

  8. [23]

    arXiv preprint arXiv:2409.17892

    EMMA-500: Enhancing Massively Multilingual Adaptation of Large Language Models . arXiv preprint arXiv:2409.17892. Lin, Peiqin, Shaoxiong Ji, Jörg Tiedemann, André FT Martins, and Hinrich Schütze

  9. [24]

    arXiv preprint arXiv:2401.13303

    MaLA-500: Massive Language Adaptation of Large Language Models. arXiv preprint arXiv:2401.13303. Downey, C. M., Terra Blevins, Dhwani Serai, Dwija Parikh, and Shane Steinert-Threlkeld

    arXiv

  10. [25]

    arXiv preprint arXiv:2405.12413

    Targeted Multilingual Adaptation for Low-resource Language Families . arXiv preprint arXiv:2405.12413. Teaching a Language Model to Speak the Language of Tools 19 Huang, Shijue, Wanjun Zhong, Jianqiao Lu, Qi Zhu, Jiahui Gao, Weiwen Liu, and Yutai Hou

    arXiv

  11. [26]

    arXiv preprint arXiv:2401.17167

    Planning, Creation, Usage: Benchmarking LLMs for Comprehensive Tool Utilization in Real-World Complex Scenarios . arXiv preprint arXiv:2401.17167. Huang, Yue, Jiawen Shi, Yuan Li, Chenrui Fan, Siyuan Wu, Qihui Zhang, and Yixin Liu

    arXiv

  12. [27]

    arXiv preprint arXiv:2310.03128

    MetaTool Benchmark for Large Language Models: Deciding Whether to Use Tools and Which to Use . arXiv preprint arXiv:2310.03128. Chen, Zehui, Weihua Du, Wenwei Zhang, Kuikun Liu, Jiangning Liu, Miao Zheng, and Jingming Zhuo

    arXiv

  13. [28]

    arXiv preprint arXiv:2312.14033

    T-Eval: Evaluating the Tool Utilization Capability of Large Language Models Step by Step . arXiv preprint arXiv:2312.14033. Chen, Po-Heng, and Yun-Nung Chen

    arXiv

  14. [29]

    Efficient Unseen Language Adaptation for Multilingual Pre-Trained Language Models

    "Efficient Unseen Language Adaptation for Multilingual Pre-Trained Language Models." Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 18983-18994. Shi, Haizhou, Zihao Xu, Hengyi Wang, Weiyi Qin, Wenyuan Wang, Yibin Wang, Zifeng Wang, Sayn...

    2024

  15. [30]

    arXiv preprint arXiv:2408.04682

    ToolSandbox: A Stateful, Conversational, Interactive Evaluation Benchmark for LLM Tool Use Capabilities. arXiv preprint arXiv:2408.04682. Alexandrov, Anton, Veselin Raychev, Dimitar I. Dimitrov, Ce Zhang, Martin Vechev, and Kristina Toutanova

    arXiv

  16. [31]

    arXiv preprint arXiv:2412.10893

    BgGPT 1.0: Extending English-centric LLMs to other languages. arXiv preprint arXiv:2412.10893. Qin, Yujia, Shihao Liang, Yining Ye, Kunlun Zhu, Lan Yan, Yaxi Lu, and Yankai Lin

    arXiv

  17. [32]

    arXiv:2307.16789

    ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs. arXiv:2307.16789. Teaching a Language Model to Speak the Language of Tools 20 Appendix A Note on Naming Convention: TUCAN refers to the model series acronym (Tool-Using Capable Assistant Navigator), ...

    arXiv

  18. [2018]

    arXiv preprint arXiv:1803.05457

    Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge. arXiv preprint arXiv:1803.05457. Hu, Edward J.,, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen

    arXiv

  19. [2019]

    Zhao, Jun, Zhihao Zhang, Luhui Gao, Qi Zhang, Tao Gui, and Xuanjing Huang

    Hellaswag: Can a machine really finish your sentence? arXiv preprint arXiv:1905.07830. Zhao, Jun, Zhihao Zhang, Luhui Gao, Qi Zhang, Tao Gui, and Xuanjing Huang

    1905 arXiv

  20. [2020]

    Transformers: State-of-the-art natural language processing

    "Transformers: State-of-the-art natural language processing." Proceedings of the 2020 conference on empirical methods in natural language processing: system demonstrations, pp. 38-45. Teaching a Language Model to Speak the Language of Tools 18 Zellers, Rowan, Ari Holtzman, Yon...

    2020

  21. [2022]

    Recent work demonstrates that soft-prompt tuning can achieve effective cross-lingual transfer using merely 0.28% of tuned parameters (Chen and Chen 2024)

    enables substantial capability improvements while updating only a small fraction of model parameters. Recent work demonstrates that soft-prompt tuning can achieve effective cross-lingual transfer using merely 0.28% of tuned parameters (Chen and Chen 2024). The challenge of cat...

    2024

  22. [2023]

    arXiv preprint arXiv:2306.05301

    Toolalpaca: Generalized tool learning for language models with 3000 simulated cases. arXiv preprint arXiv:2306.05301. Joshi, Raviraj, Kanishk Singla, Anusha Kamath, Raunak Kalani, Rakesh Paul, Utkarsh Vaidya, Sanjay Singh Chauhan, Niranjan Wartikar, and Eileen Long

    arXiv

  23. [2024]

    book a flight to Barcelona

    (Joshi, et al. 2024). A critical capability gap persists across multilingual contexts, where even models with strong comprehension abilities falter when tasked with tool integration. Consider state-of-the-art Bulgarian language models like BgGPT (Alexandrov, et al. 2024): whil...

    2024

  24. [2025]

    curse of multilinguality

    demonstrate remarkable multilingual capabilities while serving as a starting point for language-specific adaptations. A critical issue in multilingual modeling is addressing the “curse of multilinguality” (Conneau, et al. 2019), where massively multilingual training dilutes pe...

    2019

Pith tools

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