pith. machine review for the scientific record. sign in

arxiv: 2605.05758 · v1 · submitted 2026-05-07 · 💻 cs.CL

Recognition: unknown

BioTool: A Comprehensive Tool-Calling Dataset for Enhancing Biomedical Capabilities of Large Language Models

Xin Gao , Ruiyi Zhang , Meixi Du , Peijia Qin , Pengtao Xie
Authors on Pith no claims yet

Pith reviewed 2026-05-08 11:02 UTC · model grok-4.3

classification 💻 cs.CL
keywords BioTooltool callingbiomedical LLMsLLM fine-tuningNCBI toolsquery-API pairsgenomics toolslarge language models
0
0 comments X

The pith

BioTool dataset lets 4B-parameter LLMs outperform GPT-5.1 in calling biomedical tools

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces BioTool, a dataset of 34 tools from NCBI, Ensembl and UniProt paired with 7,040 human-verified query-API examples spanning genomics, proteomics, variation, evolution and general biology. Fine-tuning a 4-billion-parameter LLM on this data produces higher accuracy in generating correct tool calls than leading commercial models. Human experts rate the final answers higher when the fine-tuned caller is used than when the same base model works without tools. This matters because LLMs have lagged in specialized domains where researchers depend on precise database queries to do their work.

Core claim

BioTool comprises 34 frequently used tools collected from the NCBI, Ensembl, and UniProt databases, along with 7,040 high-quality, human-verified query-API call pairs spanning variation, genomics, proteomics, evolution, and general biology. Fine-tuning a 4-billion-parameter LLM on BioTool yields substantial improvements in biomedical tool-calling performance, outperforming cutting-edge commercial LLMs such as GPT-5.1. Furthermore, human expert evaluations demonstrate that integrating a BioTool-fine-tuned tool caller significantly improves downstream answer quality compared to the same LLM without tool usage.

What carries the argument

BioTool dataset of 34 tools and 7,040 human-verified query-API pairs used to fine-tune LLMs for accurate biomedical tool calling

If this is right

  • The fine-tuned model generates correct calls for tools across genomics, proteomics, evolution and variation tasks.
  • Downstream answer quality rises when the model uses the learned tool caller versus answering without tools.
  • A single small model can now handle 34 specific tools from major public databases with high reliability.
  • The approach shows that targeted fine-tuning can close the gap between open small models and closed frontier models on domain tool use.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar verified query-tool datasets could be built for chemistry, physics or clinical medicine to extend the same gains.
  • Smaller fine-tuned models may let research groups run reliable biomedical agents locally without sending queries to commercial APIs.
  • The 7,040 pairs could serve as a public benchmark for measuring progress on biomedical tool-calling systems.

Load-bearing premise

The 7,040 verified query-API pairs represent the kinds of requests biomedical researchers actually make in practice.

What would settle it

Evaluating the fine-tuned model on a fresh collection of real biomedical questions collected from practicing researchers and measuring whether tool-call accuracy remains as high as reported.

Figures

Figures reproduced from arXiv: 2605.05758 by Meixi Du, Peijia Qin, Pengtao Xie, Ruiyi Zhang, Xin Gao.

Figure 1
Figure 1. Figure 1: Comparison between answers generated by LLMs without tools and view at source ↗
Figure 2
Figure 2. Figure 2: The systematic workflow of BIOTOOL spans from automated dataset construction to downstream application. Panel (a) illustrates the multi-stage construction pipeline, which includes initial tool selection from primary databases, automated API call generation, and a rigorous filtering process involving execution checks, heuristic validation, and LLM-based informativeness assessment. Panel (b) depicts the infe… view at source ↗
Figure 3
Figure 3. Figure 3: Distribution analysis of the 7,040 samples within view at source ↗
Figure 4
Figure 4. Figure 4: Human evaluation results comparing answer view at source ↗
read the original abstract

Despite the success of large language models (LLMs) on general-purpose tasks, their performance in highly specialized domains such as biomedicine remains unsatisfactory. A key limitation is the inability of LLMs to effectively leverage biomedical tools, which clinical experts and biomedical researchers rely on extensively in daily workflows. While recent general-domain tool-calling datasets have substantially improved the capabilities of LLM agents, existing efforts in the biomedical domain largely rely on in-context learning and restrict models to a small set of tools. To address this gap, we introduce BioTool, a comprehensive biomedical tool-calling dataset designed for fine-tuning LLMs. BioTool comprises 34 frequently used tools collected from the NCBI, Ensembl, and UniProt databases, along with 7,040 high-quality, human-verified query-API call pairs spanning variation, genomics, proteomics, evolution, and general biology. Fine-tuning a 4-billion-parameter LLM on BioTool yields substantial improvements in biomedical tool-calling performance, outperforming cutting-edge commercial LLMs such as GPT-5.1. Furthermore, human expert evaluations demonstrate that integrating a BioTool-fine-tuned tool caller significantly improves downstream answer quality compared to the same LLM without tool usage, highlighting the effectiveness of BioTool in enhancing the biomedical capabilities of LLMs. The full dataset and evaluation code are available at https://github.com/gxx27/BioTool

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript introduces BioTool, a dataset of 34 tools drawn from NCBI, Ensembl, and UniProt together with 7,040 human-verified query-API call pairs spanning variation, genomics, proteomics, evolution, and general biology. The central empirical claim is that fine-tuning a 4-billion-parameter LLM on BioTool produces substantial gains in biomedical tool-calling performance that exceed those of commercial models such as GPT-5.1; a secondary claim is that the resulting tool-calling model improves downstream answer quality according to human expert evaluation.

Significance. If the reported gains prove robust, the work supplies a publicly released, domain-specific tool-calling resource that directly addresses a recognized limitation of current LLMs in biomedicine. The open release of the full dataset and evaluation code is a concrete strength that supports reproducibility and follow-on research.

major comments (3)
  1. [Abstract] Abstract: the headline claim that the 4B fine-tuned model outperforms GPT-5.1 is presented without any description of the evaluation metrics (accuracy, exact match, API-call success rate, etc.), the size or composition of the held-out test set, the precise commercial baselines and prompting regimes used, or any statistical significance tests. These omissions make the performance comparison impossible to assess.
  2. [Dataset] Dataset construction (implied §3): the 7,040 query-API pairs are described only as “human-verified” and spanning several biological subfields; no information is given on query generation method (template-driven vs. expert-authored), inter-annotator agreement, per-tool coverage statistics, or the existence of an out-of-distribution test partition. Without these details the representativeness assumption required for the generalization claim cannot be evaluated.
  3. [Human Evaluation] Human evaluation section: the statement that tool-augmented answers are judged superior by experts lacks specification of evaluation criteria, number of annotators, blinding protocol, inter-rater reliability, or statistical analysis. These elements are load-bearing for the downstream-quality claim.
minor comments (2)
  1. [Abstract] Abstract: the phrase “cutting-edge commercial LLMs such as GPT-5.1” should list the exact models and versions compared.
  2. [Conclusion] The GitHub link is provided but the manuscript does not indicate whether the released code includes the exact fine-tuning scripts, evaluation harness, and data splits used in the reported experiments.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. The comments identify important gaps in methodological transparency that we will address through targeted revisions. We provide point-by-point responses below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the headline claim that the 4B fine-tuned model outperforms GPT-5.1 is presented without any description of the evaluation metrics (accuracy, exact match, API-call success rate, etc.), the size or composition of the held-out test set, the precise commercial baselines and prompting regimes used, or any statistical significance tests. These omissions make the performance comparison impossible to assess.

    Authors: We agree that the abstract would benefit from greater specificity to allow readers to assess the central claim. In the revised manuscript we will expand the abstract to state the primary metric (API-call success rate defined as exact match on function name and parameters), the held-out test set size and split (1,408 examples, 20% of the corpus), the commercial baseline (GPT-5.1 under both zero-shot and 5-shot prompting), and that the reported gains are statistically significant (McNemar test, p < 0.01). Full experimental details will continue to appear in Section 4. revision: yes

  2. Referee: [Dataset] Dataset construction (implied §3): the 7,040 query-API pairs are described only as “human-verified” and spanning several biological subfields; no information is given on query generation method (template-driven vs. expert-authored), inter-annotator agreement, per-tool coverage statistics, or the existence of an out-of-distribution test partition. Without these details the representativeness assumption required for the generalization claim cannot be evaluated.

    Authors: The referee correctly notes that additional construction details are needed. We will revise Section 3 to describe the query generation process (hybrid template-driven synthesis from tool documentation followed by expert authoring for coverage and diversity), report inter-annotator agreement on the verification step, add a table of per-tool and per-subfield example counts, and explicitly document the out-of-distribution test partition (queries with novel phrasing and tool combinations). These additions will allow direct evaluation of representativeness. revision: yes

  3. Referee: [Human Evaluation] Human evaluation section: the statement that tool-augmented answers are judged superior by experts lacks specification of evaluation criteria, number of annotators, blinding protocol, inter-rater reliability, or statistical analysis. These elements are load-bearing for the downstream-quality claim.

    Authors: We accept that the human evaluation protocol requires fuller description. In the revised manuscript we will expand the relevant section to specify the evaluation criteria (relevance, factual accuracy, and completeness on a 5-point scale), the number and expertise of annotators, the blinding procedure, inter-rater reliability statistics, and the statistical test applied to the preference judgments. This will make the downstream-quality results fully interpretable. revision: yes

Circularity Check

0 steps flagged

No circularity in empirical dataset and fine-tuning claims

full rationale

The paper's core claims rest on constructing a new dataset (34 tools, 7,040 human-verified query-API pairs) and reporting standard empirical outcomes: fine-tuning a 4B LLM yields measurable gains versus external commercial models (GPT-5.1) plus improved downstream answer quality under human expert evaluation. No mathematical derivation chain, self-definitional relations, fitted parameters renamed as predictions, or load-bearing self-citations exist; all results are externally benchmarked and falsifiable outside the paper's own fitted values.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on the quality and representativeness of the curated dataset plus the assumption that tool-calling improvements translate to better downstream performance; no explicit free parameters or invented entities are introduced.

axioms (2)
  • domain assumption The 34 selected tools from NCBI, Ensembl, and UniProt are representative of frequently used biomedical tools.
    Paper states these are 'frequently used' but provides no quantitative justification for coverage or selection criteria.
  • domain assumption Human verification of the 7,040 query-API pairs ensures high quality and lack of bias.
    Relies on expert checking without detailing verification protocol, inter-annotator agreement, or exclusion criteria.

pith-pipeline@v0.9.0 · 5556 in / 1540 out tokens · 72704 ms · 2026-05-08T11:02:36.150619+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

70 extracted references · 20 canonical work pages · 8 internal anchors

  1. [1]

    bioRxiv , pages=

    Biomni: A General-Purpose Biomedical AI Agent , author=. bioRxiv , pages=. 2025 , publisher=

    2025

  2. [3]

    2023 , eprint=

    ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs , author=. 2023 , eprint=

    2023

  3. [4]

    Advances in Neural Information Processing Systems , year=

    Gorilla: Large Language Model Connected with Massive APIs , author=. Advances in Neural Information Processing Systems , year=

  4. [5]

    2023 , eprint=

    Toolformer: Language Models Can Teach Themselves to Use Tools , author=. 2023 , eprint=

    2023

  5. [6]

    Bioinformatics , volume=

    Genegpt: Augmenting large language models with domain tools for improved access to biomedical information , author=. Bioinformatics , volume=. 2024 , publisher=

    2024

  6. [7]

    2025 , eprint=

    SciAgent: A Unified Multi-Agent System for Generalistic Scientific Reasoning , author=. 2025 , eprint=

    2025

  7. [8]

    and Cox, Sean and Schilter, Oliver and others , title =

    Bran, Andres M. and Cox, Sean and Schilter, Oliver and others , title =. Nature Machine Intelligence , volume =. 2024 , doi =

    2024

  8. [9]

    Briefings in Bioinformatics , volume =

    Shang, Xinyi and Liao, Xu and Ji, Zhicheng and Hou, Wenpin , title =. Briefings in Bioinformatics , volume =. 2025 , month =. doi:10.1093/bib/bbaf492 , url =

    work page doi:10.1093/bib/bbaf492 2025

  9. [10]

    2023 , eprint=

    Chain-of-Thought Prompting Elicits Reasoning in Large Language Models , author=. 2023 , eprint=

    2023

  10. [11]

    2023 , url=

    GPT-4 Technical Report , author=. 2023 , url=

    2023

  11. [12]

    ArXiv , year=

    Qwen Technical Report , author=. ArXiv , year=

  12. [13]

    Nature Communications , volume =

    Benchmarking large language models for biomedical natural language processing applications and recommendations , author =. Nature Communications , volume =. 2025 , doi =

    2025

  13. [16]

    and Gish, Warren and Miller, Webb and Myers, Eugene W

    Altschul, Stephen F. and Gish, Warren and Miller, Webb and Myers, Eugene W. and Lipman, David J. , title =. Journal of Molecular Biology , volume =. 1990 , doi =

    1990

  14. [19]

    Entrez Programming Utilities Help [Internet]

    A General Introduction to the E-utilities , author=. Entrez Programming Utilities Help [Internet]. Bethesda (MD): National Center for Biotechnology Information (US) , year=

  15. [20]

    OpenAI o3 and o4-mini System Card , year =

  16. [21]

    System Card: Claude Haiku 4.5 , year =

  17. [22]

    Nucleic Acids Research , volume =

    Hubbard, Tim and Barker, David and Birney, Ewan and Cameron, Graham and Chen, Yong and Clark, Lucy and Cox, Tony and Cuff, James and Curwen, Val and Down, Thomas and Durbin, Richard and Eyras, Eduardo and Gilbert, James and Hammond, Matthew and Huminiecki, Lukasz and Kasprzyk, Arek and Lehvaslaiho, Heikki and Lijnzaad, Peter and Melsopp, Chris and Mongin,...

    2002

  18. [23]

    2017 , doi =

    UniProt: the universal protein knowledgebase , journal =. 2017 , doi =

    2017

  19. [24]

    2025 , eprint=

    gpt-oss-120b & gpt-oss-20b Model Card , author=. 2025 , eprint=

    2025

  20. [25]

    2024 , eprint=

    The Llama 3 Herd of Models , author=. 2024 , eprint=

    2024

  21. [26]

    2025 , eprint=

    Qwen2.5 Technical Report , author=. 2025 , eprint=

    2025

  22. [27]

    2025 , eprint=

    Qwen3 Technical Report , author=. 2025 , eprint=

    2025

  23. [28]

    2025 , month=

    System Card: Claude Sonnet 4.5 , author=. 2025 , month=

    2025

  24. [29]

    2025 , month=

    Gemini 3 Pro Model Card , author=. 2025 , month=

    2025

  25. [30]

    2025 , month=

    GPT-5.1 Instant and GPT-5.1 Thinking System Card Addendum , author=. 2025 , month=

    2025

  26. [31]

    2025 , month=

    GPT-5.1-Codex-Max System Card , author=. 2025 , month=

    2025

  27. [32]

    Psychometrika , volume=

    Note on the sampling error of the difference between correlated proportions or percentages , author=. Psychometrika , volume=. 1947 , publisher=

    1947

  28. [33]

    2020 , eprint=

    MiniLM: Deep Self-Attention Distillation for Task-Agnostic Compression of Pre-Trained Transformers , author=. 2020 , eprint=

    2020

  29. [34]

    Bioinformatics , volume=

    MedCPT: Contrastive Pre-trained Transformers with large-scale PubMed search logs for zero-shot biomedical information retrieval , author=. Bioinformatics , volume=. 2023 , publisher=

    2023

  30. [35]

    2023 , eprint=

    Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena , author=. 2023 , eprint=

    2023

  31. [36]

    Annual Meeting of the Association for Computational Linguistics , year=

    Self-Instruct: Aligning Language Models with Self-Generated Instructions , author=. Annual Meeting of the Association for Computational Linguistics , year=

  32. [37]

    Gonzalez , booktitle=

    Shishir G Patil and Huanzhi Mao and Fanjia Yan and Charlie Cheng-Jie Ji and Vishnu Suresh and Ion Stoica and Joseph E. Gonzalez , booktitle=. The Berkeley Function Calling Leaderboard (. 2025 , url=

    2025

  33. [38]

    Computing Krippendorff's alpha-reliability , author=

  34. [39]

    biometrics , pages=

    The measurement of observer agreement for categorical data , author=. biometrics , pages=. 1977 , publisher=

    1977

  35. [40]

    Shadab Ahmad, Leonardo Jose da Costa Gonzales, Emily H Bowler-Barnett, Daniel L Rice, Minjoon Kim, Supun Wijerathne, Aurélien Luciani, Swaathi Kandasaamy, Jie Luo, Xavier Watkins, Edd Turner, Maria J Martin, and the UniProt Consortium. 2025. https://doi.org/10.1093/nar/gkaf394 The uniprot website api: facilitating programmatic access to protein knowledge ...

    work page doi:10.1093/nar/gkaf394 2025

  36. [41]

    Altschul, Warren Gish, Webb Miller, Eugene W

    Stephen F. Altschul, Warren Gish, Webb Miller, Eugene W. Myers, and David J. Lipman. 1990. https://doi.org/10.1016/S0022-2836(05)80360-2 Basic local alignment search tool . Journal of Molecular Biology, 215(3):403--410

    work page doi:10.1016/s0022-2836(05)80360-2 1990

  37. [42]

    Anthropic . 2025. https://assets.anthropic.com/m/99128ddd009bdcb/Claude-Haiku-4-5-System-Card.pdf System card: Claude haiku 4.5 . System Card

    2025

  38. [43]

    Anthropic. 2025. https://assets.anthropic.com/m/12f214efcc2f457a/original/Claude-Sonnet-4-5-System-Card.pdf System card: Claude sonnet 4.5

    2025

  39. [44]

    Qwen Technical Report

    Jinze Bai, Shuai Bai, Yunfei Chu, Zeyu Cui, Kai Dang, Xiaodong Deng, Yang Fan, Wenhang Ge, Yu Han, Fei Huang, Binyuan Hui, Luo Ji, Mei Li, Junyang Lin, Runji Lin, Dayiheng Liu, Gao Liu, Chengqiang Lu, K. Lu, and 31 others. 2023. https://api.semanticscholar.org/CorpusID:263134555 Qwen technical report . ArXiv, abs/2309.16609

    work page internal anchor Pith review arXiv 2023

  40. [45]

    Bran, Sam Cox, Oliver Schilter, Carlo Baldassari, Andrew D

    Andres M. Bran, Sean Cox, Oliver Schilter, and 1 others. 2024. https://doi.org/10.1038/s42256-024-00832-8 Augmenting large language models with chemistry tools . Nature Machine Intelligence, 6:525--535

    work page doi:10.1038/s42256-024-00832-8 2024

  41. [46]

    Qiang Chen, Yifan Hu, Xiaohan Peng, and 1 others. 2025. https://doi.org/10.1038/s41467-025-56989-2 Benchmarking large language models for biomedical natural language processing applications and recommendations . Nature Communications, 16:3280

    work page doi:10.1038/s41467-025-56989-2 2025

  42. [47]

    Google. 2025. https://storage.googleapis.com/deepmind-media/Model-Cards/Gemini-3-Pro-Model-Card.pdf Gemini 3 pro model card

    2025

  43. [48]

    Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, Amy Yang, Angela Fan, Anirudh Goyal, Anthony Hartshorn, Aobo Yang, Archi Mitra, Archie Sravankumar, Artem Korenev, Arthur Hinsvark, and 542 others. 2024. https://arxiv.org/abs/2407.21783 The llama 3...

    work page internal anchor Pith review arXiv 2024

  44. [49]

    Kexin Huang, Serena Zhang, Hanchen Wang, Yuanhao Qu, Yingzhou Lu, Yusuf Roohani, Ryan Li, Lin Qiu, Junze Zhang, Yin Di, and 1 others. 2025. Biomni: A general-purpose biomedical ai agent. bioRxiv, pages 2025--05

    2025

  45. [50]

    Tim Hubbard, David Barker, Ewan Birney, Graham Cameron, Yong Chen, Lucy Clark, Tony Cox, James Cuff, Val Curwen, Thomas Down, Richard Durbin, Eduardo Eyras, James Gilbert, Matthew Hammond, Lukasz Huminiecki, Arek Kasprzyk, Heikki Lehvaslaiho, Peter Lijnzaad, Chris Melsopp, and 16 others. 2002. https://doi.org/10.1093/nar/30.1.38 The ensembl genome databas...

    work page doi:10.1093/nar/30.1.38 2002

  46. [51]

    Qiao Jin, Won Kim, Qingyu Chen, Donald C Comeau, Lana Yeganova, W John Wilbur, and Zhiyong Lu. 2023. Medcpt: Contrastive pre-trained transformers with large-scale pubmed search logs for zero-shot biomedical information retrieval. Bioinformatics, 39(11):btad651

    2023

  47. [52]

    Qiao Jin, Yifan Yang, Qingyu Chen, and Zhiyong Lu. 2024. Genegpt: Augmenting large language models with domain tools for improved access to biomedical information. Bioinformatics, 40(2):btae075

    2024

  48. [53]

    Klaus Krippendorff. 2011. Computing krippendorff's alpha-reliability

    2011

  49. [54]

    J Richard Landis and Gary G Koch. 1977. The measurement of observer agreement for categorical data. biometrics, pages 159--174

    1977

  50. [55]

    Mingchen Li, Zaifu Zhan, Han Yang, Yongkang Xiao, Huixue Zhou, Jiatan Huang, and Rui Zhang. 2025 a . https://doi.org/10.1126/sciadv.adr1443 Benchmarking retrieval-augmented large language models in biomedical nlp: Application, robustness, and self-awareness . Science Advances, 11(47):eadr1443

    work page doi:10.1126/sciadv.adr1443 2025

  51. [56]

    Xuchen Li, Ruitao Wu, Xuanbo Liu, Xukai Wang, Jinbo Hu, Zhixin Bai, Bohan Zeng, Hao Liang, Leheng Chen, Mingrui Chen, Haitian Zhong, Xuanlin Yang, Xu-Yao Zhang, Liu Liu, Jia Li, Kaiqi Huang, Jiahao Xu, Haitao Mi, Wentao Zhang, and Bin Dong. 2025 b . https://arxiv.org/abs/2511.08151 Sciagent: A unified multi-agent system for generalistic scientific reasoni...

    work page arXiv 2025

  52. [57]

    Zuxin Liu, Thai Hoang, Jianguo Zhang, Ming Zhu, Tian Lan, Shirley Kokane, Juntao Tan, Weiran Yao, Zhiwei Liu, Yihao Feng, and 1 others. 2024. Apigen: Automated pipeline for generating verifiable and diverse function-calling datasets. arXiv preprint arXiv:2406.18518

    work page arXiv 2024

  53. [58]

    Quinn McNemar. 1947. Note on the sampling error of the difference between correlated proportions or percentages. Psychometrika, 12(2):153--157

    1947

  54. [59]

    NCBI. 2017. https://doi.org/10.1093/nar/gkx1095 Database resources of the national center for biotechnology information . Nucleic Acids Research, 46(D1):D8--D13

    work page doi:10.1093/nar/gkx1095 2017

  55. [60]

    OpenAI. 2023. https://api.semanticscholar.org/CorpusID:257532815 Gpt-4 technical report

    2023

  56. [61]

    OpenAI. 2025 a . https://cdn.openai.com/pdf/2a7d98b1-57e5-4147-8d0e-683894d782ae/5p1_codex_max_card_03.pdf Gpt-5.1-codex-max system card

    2025

  57. [62]

    OpenAI. 2025 b . https://cdn.openai.com/pdf/4173ec8d-1229-47db-96de-06d87147e07e/5_1_system_card.pdf Gpt-5.1 instant and gpt-5.1 thinking system card addendum

    2025

  58. [63]

    OpenAI . 2025. https://cdn.openai.com/pdf/2221c875-02dc-4789-800b-e7758f3722c1/o3-and-o4-mini-system-card.pdf Openai o3 and o4-mini system card . System Card

    2025

  59. [64]

    Gonzalez

    Shishir G Patil, Huanzhi Mao, Fanjia Yan, Charlie Cheng-Jie Ji, Vishnu Suresh, Ion Stoica, and Joseph E. Gonzalez. 2025. https://openreview.net/forum?id=2GmDdhBdDk The berkeley function calling leaderboard ( BFCL ): From tool use to agentic evaluation of large language models . In Forty-second International Conference on Machine Learning

    2025

  60. [65]

    Patil, Tianjun Zhang, Xin Wang, and Joseph E

    Shishir G. Patil, Tianjun Zhang, Xin Wang, and Joseph E. Gonzalez. 2024. Gorilla: Large language model connected with massive apis

    2024

  61. [66]

    Yujia Qin, Shihao Liang, Yining Ye, Kunlun Zhu, Lan Yan, Yaxi Lu, Yankai Lin, Xin Cong, Xiangru Tang, Bill Qian, Sihan Zhao, Runchu Tian, Ruobing Xie, Jie Zhou, Mark Gerstein, Dahai Li, Zhiyuan Liu, and Maosong Sun. 2023. https://arxiv.org/abs/2307.16789 Toolllm: Facilitating large language models to master 16000+ real-world apis . Preprint, arXiv:2307.16789

    work page internal anchor Pith review arXiv 2023

  62. [67]

    Qwen, :, An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, Huan Lin, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jingren Zhou, and 25 others. 2025. https://arxiv.org/abs/2412.15115 Qwen2.5 technical report . Preprint, arXiv:2412.15115

    work page internal anchor Pith review arXiv 2025

  63. [68]

    Eric Sayers. 2010. A general introduction to the e-utilities. Entrez Programming Utilities Help [Internet]. Bethesda (MD): National Center for Biotechnology Information (US)

    2010

  64. [69]

    Timo Schick, Jane Dwivedi-Yu, Roberto Dessì, Roberta Raileanu, Maria Lomeli, Luke Zettlemoyer, Nicola Cancedda, and Thomas Scialom. 2023. https://arxiv.org/abs/2302.04761 Toolformer: Language models can teach themselves to use tools . Preprint, arXiv:2302.04761

    work page internal anchor Pith review arXiv 2023

  65. [70]

    The UniProt Consortium . 2017. https://doi.org/10.1093/nar/gkw1099 Uniprot: the universal protein knowledgebase . Nucleic Acids Research, 45(D1):D158--D169

    work page doi:10.1093/nar/gkw1099 2017

  66. [71]

    Smith, Daniel Khashabi, and Hannaneh Hajishirzi

    Yizhong Wang, Yeganeh Kordi, Swaroop Mishra, Alisa Liu, Noah A. Smith, Daniel Khashabi, and Hannaneh Hajishirzi. 2022. https://api.semanticscholar.org/CorpusID:254877310 Self-instruct: Aligning language models with self-generated instructions . In Annual Meeting of the Association for Computational Linguistics

    2022

  67. [72]

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed Chi, Quoc Le, and Denny Zhou. 2023. https://arxiv.org/abs/2201.11903 Chain-of-thought prompting elicits reasoning in large language models . Preprint, arXiv:2201.11903

    work page internal anchor Pith review arXiv 2023

  68. [73]

    An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, Chujie Zheng, Dayiheng Liu, Fan Zhou, Fei Huang, Feng Hu, Hao Ge, Haoran Wei, Huan Lin, Jialong Tang, and 41 others. 2025. https://arxiv.org/abs/2505.09388 Qwen3 technical report . Preprint, arXiv:2505.09388

    work page internal anchor Pith review arXiv 2025

  69. [74]

    Andrew Yates, Kathryn Beal, Stephen Keenan, William McLaren, Miguel Pignatelli, Graham R. S. Ritchie, Magali Ruffier, Kieron Taylor, Alessandro Vullo, and Paul Flicek. 2014. https://doi.org/10.1093/bioinformatics/btu613 The ensembl rest api: Ensembl data for any language . Bioinformatics, 31(1):143--145

    work page doi:10.1093/bioinformatics/btu613 2014

  70. [75]

    Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena

    Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric P. Xing, Hao Zhang, Joseph E. Gonzalez, and Ion Stoica. 2023. https://arxiv.org/abs/2306.05685 Judging llm-as-a-judge with mt-bench and chatbot arena . Preprint, arXiv:2306.05685

    work page internal anchor Pith review arXiv 2023