Pith. sign in

REVIEW 5 major objections 5 minor 56 references

ToolLIFT: Lifting Tool-Specific Trajectories into Function-Level Graphs for Generalizable Tool Planning

T0 review · 5 major / 5 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 The paper argues that tool-use experience transfers between tool sets when stored as function-level workflow graphs rather than tool-level graphs, and reports consistent gains on out-of-distribution benchmarks.

desk verdict Function-level workflow graphs are a genuinely useful abstraction for tool planning, and the paper's OOD gains are plausible, but the central transfer mechanism is never directly validated and the evaluation has enough single-run/filtered-set issues that the margins should be treated as provisional. read the letter →

arxiv 2608.03468 v1 pith:F2646LSB submitted 2026-08-04 cs.AI

Xiuhui You , Jiayi Luo , Zichao Shen , Qingyun Sun , Ziwei Zhang This is my paper
classification cs.AI
keywords toolplanningLLMagentsfunction-levelworkflowgraphout-of-distributiongeneralizationtrajectoryliftingreinforcementlearningsourcetracingclustering
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

Tool-use experience is usually stored as graphs whose nodes are concrete tools, so a plan learned for one tool set is useless for another. ToolLIFT's central claim is that the reusable part of a trajectory is its function-level workflow: clustering tools by what they do rather than what domain they serve lets historical transitions be aggregated into a Function-Level Workflow Graph (FWG). The framework then plans in two steps—first choose the function-level workflow using the FWG as soft guidance, then instantiate each function with a concrete tool—and uses reinforcement learning to make each argument explicitly a direct value or a reference to a prior tool output. On two in-distribution and three out-of-distribution benchmarks, ToolLIFT outperforms existing planners, with the largest margins on tool sets unseen during training (up to 4.90 accuracy points over the strongest baseline). If the claim holds, an agent can be extended to new tools by reading their schemas and inheriting workflow experience from functionally equivalent tools.

What carries the argument

The central object is the Function-Level Workflow Graph (FWG): a directed, weighted graph whose nodes are functional clusters of tools and whose edge weights are the probabilities of one function following another in historical trajectories. It is built by trajectory lifting—mapping each concrete tool to a functional cluster via LLM decomposition, embedding, UMAP projection, and K-means—and it is what makes experience portable. The FWG does three jobs: it shares collaboration evidence across all tools in a cluster, it supplies soft transition guidance for global workflow planning, and, through nearest-centroid assignment, it lets an unseen tool inherit a function's transitions (cold-start in

What would settle it

Take an unseen tool and run its schema through the pipeline twice, once with the original functional description and once with a synonym-heavy paraphrase that emphasizes domain details; if the two near-identical tools land in different clusters, the nearest-centroid assignment is not stable. A stronger version: rebuild the FWG with random tool-to-cluster assignments of the same size and re-run the OOD benchmarks; if accuracy stays close to the full model, the function abstraction is not what carries the generalization.

Watch

Extended reading notes

Core claim

The paper claims that analogous tasks performed with different concrete tools share a common function-level workflow structure, and that this structure is the right unit for transferring tool-use experience. ToolLIFT operationalizes the claim by decomposing each tool schema into a functional description and a domain description, embedding only the functional part, projecting the embeddings, and clustering tools into functional roles. Historical tool trajectories are then lifted to sequences of functional roles, and adjacent-role pairs are counted into transition probabilities that form the FWG. An unseen tool is assigned to the role of the nearest cluster centroid and inherits that role's tr

Load-bearing premise

An unseen tool is assigned to whichever functional group it is closest to and inherits that group's workflow experience; if the grouping is wrong, the experience it inherits describes a different job, and the claimed transfer fails.

Editorial extensions

If this is right

  • Trajectory libraries collected for one tool set can guide planning with a completely disjoint tool set, requiring only schema descriptions for the new tools.
  • Planning the full function-level workflow before choosing tools prevents locally plausible single-tool choices from breaking the overall plan, as the paper's case study shows.
  • Explicitly marking each argument as direct or as a reference to a prior output reduces source errors compared with a reward that only exact-matches values.
  • Rarely used tools benefit most from function-level sharing, because transitions learned from abundant functionally equivalent tools fill in for sparse tool-level evidence.
  • Because gains appear with two different open LLM backbones, the function-level abstraction itself, not a specific model, is the driver.

Reading between the lines

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

  • The FWG could be used as a cheap pre-filter for tool retrieval: function-level transitions can narrow the candidate tool set before an LLM is asked to plan, which would matter for agents with thousands of APIs. This is an extension, not a claim in the paper.
  • Cluster coherence is a measurable predictor of transferability: one could check whether the same functional roles co-cluster across different trajectory corpora, and use that overlap to forecast OOD gains before running a planner.
  • The single-source argument assumption is a natural boundary condition; a testable extension is to let one argument combine several prior outputs and extend the source-gated reward accordingly.
  • A practical on-boarding protocol follows: when adding a tool, verify its nearest-centroid assignment against a human label of its function before trusting the inherited workflow, since the whole transfer mechanism rests on that assignment.
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

5 major / 5 minor

Summary. The paper proposes ToolLIFT, a framework for tool-use planning that abstracts tool-specific trajectories into a function-level workflow graph (FWG). Tool schemas are decomposed into functional and domain descriptions by an LLM, embedded with BGE-M3, reduced by UMAP, and clustered by K-means to define a tool-to-function mapping phi (Eq. 4). Historical trajectories are lifted to function sequences, transition counts are aggregated into an FWG, and unseen tools inherit transitions through nearest-centroid cluster assignment. Planning is decoupled into FWG-guided workflow generation and workflow-constrained tool selection, with GRPO training and source-gated, skill-specific rewards for argument-level dataflow. Experiments on two ID and three OOD benchmarks compare against five baselines with two LLM backbones, plus ablations, sensitivity analyses, and case studies.

Significance. The central idea is timely and plausible: function-level abstractions may transfer collaboration structure across disjoint tool sets, and the paper gives a concrete mechanism with clearly stated equations. If validated, the FWG would be a useful contribution to tool planning. The authors also provide useful ablations and qualitative case studies. However, the current evidence does not yet support the abstract's strong claim of 'consistently outperforming' baselines and 'strong generalization to unseen tool sets.' The evaluation is weakened by filtered test sets, single-run results without error bars, test-set hyperparameter selection, and a lack of quantitative validation of the cold-start cluster assignment that underlies the entire transfer mechanism. The manuscript is within scope and the issues are addressable, but they are load-bearing for the main claim.

major comments (5)
  1. [Sec. 4.1, Eq. (4); Appendix C.2]
  2. [Sec. B.1, Table 5]
  3. [Sec. B.6, Table 1]
  4. [Sec. C.1, Tables 1 and 3]
  5. [Sec. A.1, Eqs. (7) and (15); Table 1]
minor comments (5)
  1. [Sec. 4.2, Eq. (7)]
  2. [Sec. 5.3, Fig. 5]
  3. [Sec. B.2, Eq. (23)]
  4. [Appendix C.4]
  5. [Sec. 4.4]

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: ToolLIFT's FWG and rewards are built from held-out training trajectories and evaluated against externally defined ground-truth tool calls; the cluster mapping is a learned abstraction, not a self-fulfilling prediction.

full rationale

The derivation chain is self-contained. The FWG is constructed entirely from training-trajectory counts after a fixed, pre-specified clustering of tool schemas (BGE-M3 + UMAP + K-means, Eqs. 1–3). Unseen tools are mapped to clusters by nearest centroid (Eq. 4); this is an extrapolation assumption that could fail empirically, but it is not circular—it does not use the test labels to define the clusters or the graph. The Stage 1 workflow target W*_c is obtained by applying the same phi mapping to ground-truth tool calls (Eq. 15), which is a transformed version of external annotations, not an output of the model or of the FWG transitions. The Stage 1 reward (Eq. 7) trains the planner to recover that transformed target; the Stage 2 generator is then evaluated against ground-truth tool names and argument values (Section B.2). Thus, the final Acc/n-F1/l-F1 numbers are not forced by the clustering or by the FWG by construction. The only self-citations (the survey Wang et al. 2024b, ToolExpNet Zhang et al. 2025) appear in related work and are not used to justify the central transfer mechanism. The lack of quantitative OOD cluster-purity validation is a correctness/robustness limitation, not evidence of circularity, and the identical Seal-Tools Acc for both backbones is a potential reporting issue, not a circular reduction. The stated limitation about single-source arguments is orthogonal. Overall, the central claims are independently grounded in held-out, externally defined benchmarks.

Assumptions & free parameters 5 free parameters · 6 assumptions · 1 invented entities

The central contribution rests on the function abstraction and the clustering that defines it. The free parameters are mostly standard RL and clustering hyperparameters, but epsilon_pert is tuned on the evaluation benchmarks themselves, which is a genuine source of optimism in the reported numbers.

free parameters (5)
  • Number of functional clusters L = 30
    Selected via silhouette coefficient (Section 5.1); sensitivity analyzed in Section 5.3, Figure 5.
  • Reward scale rho = 3
    Hand-set in Section 5.1; scales workflow and correctness rewards.
  • Reward weights lambda_match, lambda_value = 1, 2
    Hand-set in Section 5.1; balance tool matching and argument value rewards.
  • Workflow perturbation probability epsilon_pert = 0.2
    Selected via sensitivity analysis in Appendix C.1 on DailyLifeAPIs and Multimedia, which are also main evaluation datasets.
  • UMAP hyperparameters (n_neighbors=15, n_components=32, min_dist=0.1) = hand-selected
    Configuration in Table 6; determine the embedding geometry that defines functional clusters.
assumptions (6)
  • domain assumption Analogous tasks share a common function-level workflow structure despite concrete tool differences.
    Motivating insight in Section 1 and Figure 1; if false, the FWG abstraction has no transferable content.
  • domain assumption An LLM prompt can reliably separate a tool schema into functional and domain descriptions.
    Used in Section 4.1 to decompose schemas before embedding; no validation of decomposition quality is reported.
  • domain assumption K-means clustering on UMAP embeddings of functional descriptions yields semantically coherent functional roles that generalize to unseen tools.
    Cold-start inheritance in Eq. 4 maps unseen tools to nearest centroids; if clusters are unstable, inherited transitions are noise.
  • domain assumption Each tool-call argument has a single source, either the query context or one preceding tool output.
    Stated in Problem Definition and acknowledged as a limitation in the Conclusion; excludes multi-source arguments.
  • standard math K-means, UMAP, and BGE-M3 behave as standard for this text-clustering task.
    Invoked in Section 4.1 with default configurations; no dedicated validation of embedding quality for tool schemas.
  • domain assumption The historical trajectory corpus is sufficiently representative to estimate function transition probabilities.
    Transition counts in Eqs. 1-3 are taken as ground truth for workflow guidance; sparse or biased trajectories would distort the FWG.
invented entities (1)
  • Function-Level Workflow Graph (FWG) independent evidence
    purpose: Transfer collaboration experience across tool sets by representing trajectories as sequences of functional roles.
    The OOD benchmarks provide a falsifiable test: if the FWG abstraction were incorrect, OOD planning would not improve; the reported gains support the construct, though not independently replicated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ToolLIFT: Lifting Tool-Specific Trajectories into Function-Level Graphs for Generalizable Tool Planning." pith.science (2026). https://pith.science/paper/F2646LSB

@misc{pith2026260803468,
  author       = {Pith},
  title        = {Pith review of: ToolLIFT: Lifting Tool-Specific Trajectories into Function-Level Graphs for Generalizable Tool Planning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/F2646LSB}},
  note         = {Machine review of arXiv:2608.03468}
}
read the original abstract

Historical tool-use trajectories provide valuable experience for large language model (LLM) agents to plan and coordinate tool usage. Existing approaches directly construct tool-level graphs from these trajectories, but the resulting graphs remain tied to specific tools and are hard to generalize across tool sets. To tackle this challenge, we find that despite differences in the tools involved, analogous tasks often share a common function-level workflow structure, which serves as a potentially more transferable abstraction for tool planning. Based on this insight, we propose ToolLIFT, a framework that lifts tool-specific trajectories into a function-level workflow graph (FWG) for generalizable tool planning. Specifically, we first propose a trajectory-lifting mechanism that encodes workflow structures in the FWG and shares collaboration experience across tools. Then, building on the global structure of the FWG, we introduce decoupled workflow planning and tool selection to align individual tool choices with the overall workflow. Lastly, to ensure reliable tool dataflow, we adopt Reinforcement Learning (RL) and propose source-gated and skill-specific rewards to maintain source-traceable information flow across tool calls. Experiments on two in-distribution (ID) and three out-of-distribution (OOD) benchmarks show that ToolLIFT consistently outperforms state-of-the-art baselines, demonstrating strong generalization to unseen tool sets.

Figures

Figures reproduced from arXiv: 2608.03468 by the authors.

Figure 1
Figure 1. Tool-use trajectories instantiated with different tool [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. An overview of ToolLIFT. (a) We lift tool-specific trajectories into a Function-Level Workflow Graph (FWG) that captures function-level collaboration structures shared across tool sets. (b) We plan a function-level workflow over the global structure of the FWG and instantiate it with concrete tools under constraints, keeping individual tool choices aligned with the overall workflow. (c) We explicitly assign argument… view at source ↗
Figure 3
Figure 3. Performance across historical tool-use frequencies [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Accuracy by tool-chain length on Multimedia (ID) [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Average accuracy under different numbers of func [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Sensitivity to the workflow-perturbation probability [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]
Figure 7
Figure 7. Figure 7: Sensitivity to the workflow-perturbation probability [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]
Figure 8
Figure 8. Figure 8: Qualitative visualization of tool functional cluster [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 9
Figure 9. Figure 9: Stage 1 GRPO reward curve [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

56 extracted references · 27 canonical work pages

  1. [1]

    Toolformer: Language Models Can Teach Themselves to Use Tools , booktitle =

    Timo Schick and Jane Dwivedi-Yu and Roberto Dess. Toolformer: Language Models Can Teach Themselves to Use Tools , booktitle =

    work page

  2. [2]

    International Conference on Learning Representations , volume=

    Toolllm: Facilitating large language models to master 16000+ real-world apis , author=. International Conference on Learning Representations , volume=

    work page

  3. [3]

    11th International Conference on Learning Representations, ICLR 2023 , year=

    REACT: SYNERGIZING REASONING AND ACTING IN LANGUAGE MODELS , author=. 11th International Conference on Learning Representations, ICLR 2023 , year=

    2023

  4. [4]

    Advances in Neural Information Processing Systems , volume=

    Hugginggpt: Solving ai tasks with chatgpt and its friends in hugging face , author=. Advances in Neural Information Processing Systems , volume=

  5. [5]

    Advances in Neural Information Processing Systems , volume=

    Gorilla: Large language model connected with massive apis , author=. Advances in Neural Information Processing Systems , volume=

  6. [6]

    Frontiers of Computer Science , volume=

    Tool learning with large language models: A survey , author=. Frontiers of Computer Science , volume=. 2025 , publisher=

    2025

  7. [7]

    European Conference on Computer Vision , pages =

    Zhaoyang Liu and Zeqiang Lai and Zhangwei Gao and Erfei Cui and Ziheng Li and Xizhou Zhu and Lewei Lu and Qifeng Chen and Yu Qiao and Jifeng Dai and Wenhai Wang , title =. European Conference on Computer Vision , pages =

    work page

  8. [8]

    Advances in Neural Information Processing Systems , volume=

    Taskbench: Benchmarking large language models for task automation , author=. Advances in Neural Information Processing Systems , volume=

Show all 56 references
  1. [9]

    arXiv preprint arXiv:2403.00839 , year=

    Toolnet: Connecting large language models with massive tools via tool graph , author=. arXiv preprint arXiv:2403.00839 , year=

    arXiv

  2. [10]

    Chen, Wenjie and Yao, Di and Li, Wenbin and Meng, Xuying and Gong, Chang and Bi, Jingping , booktitle=

  3. [11]

    Jiang, Yan and Zhou, Hao and Gu, Lizhong and Li, Tianlong and Jin, Ruinan and Zhou, Wanqi and Han, Ai , booktitle=

  4. [12]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    AutoTool: Efficient tool selection for large language model agents , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  5. [13]

    Findings of the Association for Computational Linguistics: ACL 2025 , pages=

    Toolexpnet: Optimizing multi-tool selection in llms with similarity and dependency-aware experience networks , author=. Findings of the Association for Computational Linguistics: ACL 2025 , pages=

    2025

  6. [14]

    Advances in neural information processing systems , volume=

    Reflexion: Language agents with verbal reinforcement learning , author=. Advances in neural information processing systems , volume=

  7. [15]

    Frontiers of Computer Science , volume=

    A survey on large language model based autonomous agents , author=. Frontiers of Computer Science , volume=. 2024 , publisher=

    2024

  8. [16]

    Proceedings of the 36th annual acm symposium on user interface software and technology , pages=

    Generative agents: Interactive simulacra of human behavior , author=. Proceedings of the 36th annual acm symposium on user interface software and technology , pages=

  9. [17]

    Advances in Neural Information Processing Systems , volume=

    Qian, Cheng and Acikgoz, Emre Can and He, Qi and Wang, Hongru and Chen, Xiusi and Hakkani-T. Advances in Neural Information Processing Systems , volume=

  10. [18]

    Findings of the Association for Computational Linguistics: EMNLP 2025 , pages=

    Tool Zero: Training Tool-Augmented LLMs via Pure RL from Scratch , author=. Findings of the Association for Computational Linguistics: EMNLP 2025 , pages=

    2025

  11. [19]

    International Conference on Learning Representations , volume=

    Tool-planner: Task planning with clusters across multiple tools , author=. International Conference on Learning Representations , volume=

  12. [20]

    Li, Sijia and Huang, Yuchen and Liu, Zifan and Li, Zijian and Fu, Jingjing and Song, Lei and Bian, Jiang and Zhang, Jun and Wang, Rui , journal=

  13. [21]

    Shao, Zhihong and Wang, Peiyi and Zhu, Qihao and Xu, Runxin and Song, Junxiao and Bi, Xiao and Zhang, Haowei and Zhang, Mingchuan and Li, Y. K. and Wu, Y. and Guo, Daya , journal=

  14. [22]

    Findings of the association for computational linguistics: ACL 2024 , pages=

    M3-embedding: Multi-linguality, multi-functionality, multi-granularity text embeddings through self-knowledge distillation , author=. Findings of the association for computational linguistics: ACL 2024 , pages=

    2024

  15. [23]

    Journal of computational and applied mathematics , volume=

    Silhouettes: a graphical aid to the interpretation and validation of cluster analysis , author=. Journal of computational and applied mathematics , volume=. 1987 , publisher=

    1987

  16. [24]

    , title=

    MacQueen, J. , title=. Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability , volume=

  17. [25]

    2025 , doi=

    Guo, Daya and others , journal=. 2025 , doi=

    2025

  18. [26]

    arXiv preprint arXiv:1707.06347 , year=

    Proximal Policy Optimization Algorithms , author=. arXiv preprint arXiv:1707.06347 , year=

    arXiv

  19. [27]

    Advances in neural information processing systems , volume=

    Training language models to follow instructions with human feedback , author=. Advances in neural information processing systems , volume=

  20. [28]

    2026 , howpublished =

    2026

  21. [29]

    Nature Reviews Methods Primers , volume=

    Uniform manifold approximation and projection , author=. Nature Reviews Methods Primers , volume=. 2024 , publisher=

    2024

  22. [30]

    Let Your Graph Do the Talking: Encoding Structured Data for

    Perozzi, Bryan and Fatemi, Bahare and Zelle, Dustin and Tsitsulin, Anton and Kazemi, Mehran and Al-Rfou, Rami and Halcrow, Jonathan , journal=. Let Your Graph Do the Talking: Encoding Structured Data for

  23. [31]

    Text summarization branches out , pages=

    Rouge: A package for automatic evaluation of summaries , author=. Text summarization branches out , pages=

  24. [32]

    arXiv preprint arXiv:2306.05301 , year=

    Toolalpaca: Generalized tool learning for language models with 3000 simulated cases , author=. arXiv preprint arXiv:2306.05301 , year=

    arXiv

  25. [33]

    arXiv preprint arXiv:2306.06624 , year=

    Restgpt: Connecting large language models with real-world restful apis , author=. arXiv preprint arXiv:2306.06624 , year=

    arXiv

  26. [34]

    International conference on machine learning , pages=

    Language models as zero-shot planners: Extracting actionable knowledge for embodied agents , author=. International conference on machine learning , pages=. 2022 , organization=

    2022

  27. [35]

    Advances in neural information processing systems , volume=

    Chain-of-thought prompting elicits reasoning in large language models , author=. Advances in neural information processing systems , volume=

  28. [36]

    Proceedings of the 61st annual meeting of the association for computational linguistics (volume 1: long papers) , pages=

    Plan-and-solve prompting: Improving zero-shot chain-of-thought reasoning by large language models , author=. Proceedings of the 61st annual meeting of the association for computational linguistics (volume 1: long papers) , pages=

  29. [37]

    CCF International Conference on Natural Language Processing and Chinese Computing , pages=

    Seal-tools: Self-instruct tool learning dataset for agent tuning and detailed benchmark , author=. CCF International Conference on Natural Language Processing and Chinese Computing , pages=. 2024 , organization=

    2024

  30. [38]

    Advances in Neural Information Processing Systems , volume=

    Can graph learning improve planning in llm-based agents? , author=. Advances in Neural Information Processing Systems , volume=

  31. [39]

    2024 , eprint =

    An Yang and Baosong Yang and Binyuan Hui and Bo Zheng and Bowen Yu and Chang Zhou and Chengpeng Li and Chengyuan Li and Dayiheng Liu and Fei Huang and others , title =. 2024 , eprint =

    2024

  32. [40]

    arXiv preprint arXiv:2407.21783 , year=

    The llama 3 herd of models , author=. arXiv preprint arXiv:2407.21783 , year=

    arXiv

  33. [41]

    Zhao, Andrew and Huang, Daniel and Xu, Quentin and Lin, Matthieu and Liu, Yong-Jin and Huang, Gao , journal =

  34. [42]

    Advances in Neural Information Processing Systems , volume=

    Autoguide: Automated generation and selection of context-aware guidelines for large language model agents , author=. Advances in Neural Information Processing Systems , volume=

  35. [43]

    International Conference on Machine Learning , pages=

    Agent Workflow Memory , author=. International Conference on Machine Learning , pages=

  36. [44]

    Wang, Chenxi and Yu, Zhuoyun and Xie, Xin and Yao, Wuguannan and Fang, Runnan and Qiao, Shuofei and Cao, Kexin and Zheng, Guozhou and Qi, Xiang and Zhang, Peng and Deng, Shumin , booktitle =

  37. [45]

    International Conference on Machine Learning , pages=

    An LLM Compiler for Parallel Function Calling , author=. International Conference on Machine Learning , pages=. 2024 , organization=

    2024

  38. [46]

    Proceedings of the 34th ACM International Conference on Information and Knowledge Management , pages=

    StepTool: Enhancing Multi-Step Tool Usage in LLMs via Step-Grained Reinforcement Learning , author=. Proceedings of the 34th ACM International Conference on Information and Knowledge Management , pages=

  39. [47]

    arXiv preprint arXiv:2505.16410 , year =

    Tool-Star: Empowering LLM-Brained Multi-Tool Reasoner via Reinforcement Learning , author =. arXiv preprint arXiv:2505.16410 , year =

    arXiv

  40. [48]

    Xu, Binfeng and Peng, Zhiyuan and Lei, Bowen and Mukherjee, Subhabrata and Liu, Yuchen and Xu, Dongkuan , journal =

  41. [49]

    Advances in neural information processing systems , volume=

    Retrieval-augmented generation for knowledge-intensive nlp tasks , author=. Advances in neural information processing systems , volume=

  42. [50]

    arXiv preprint arXiv:2504.11536 , year=

    Retool: Reinforcement learning for strategic tool use in llms , author=. arXiv preprint arXiv:2504.11536 , year=

    arXiv

  43. [51]

    Transactions on Machine Learning Research , issn =

    Voyager: An Open-Ended Embodied Agent with Large Language Models , author =. Transactions on Machine Learning Research , issn =

  44. [52]

    International conference on learning representations , volume=

    Toolace: Winning the points of llm function calling , author=. International conference on learning representations , volume=

  45. [53]

    Advances in neural information processing systems , volume=

    Adaplanner: Adaptive planning from feedback with language models , author=. Advances in neural information processing systems , volume=

  46. [54]

    and Wang, Dong and Zamani, Hamed and Han, Jiawei , booktitle =

    Jin, Bowen and Zeng, Hansi and Yue, Zhenrui and Yoon, Jinsung and Arik, Sercan O. and Wang, Dong and Zamani, Hamed and Han, Jiawei , booktitle =

  47. [55]

    and Sarkhel, Somdeb and Zhang, Chao , booktitle =

    Zhuang, Yuchen and Chen, Xiang and Yu, Tong and Mitra, Saayan and Bursztyn, Victor and Rossi, Ryan A. and Sarkhel, Somdeb and Zhang, Chao , booktitle =

  48. [56]

    and Stoica, Ion and Gonzalez, Joseph E

    Packer, Charles and Wooders, Sarah and Lin, Kevin and Fang, Vivian and Patil, Shishir G. and Stoica, Ion and Gonzalez, Joseph E. , journal =

Pith tools

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