Pith. sign in

REVIEW 3 major objections 6 minor 4 cited by

Reducing Tool Hallucination via Reliability Alignment

T0 review · 3 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Giving LLMs an explicit option to defer or switch tools cuts tool-call hallucinations by roughly 70 percent and raises reliable task completion.

desk verdict Solid, practical contribution on tool hallucination that deserves engagement; the main caveat is a real evaluator-dependence gap that revision should close. read the letter →

arxiv 2412.04141 v3 pith:TV5X63E6 submitted 2024-12-05 cs.CL

Hongshen Xu , Zichen Zhu , Lei Pan , Zihan Wang , Su Zhu , Da Ma , Ruisheng Cao , Lu Chen
show 1 more author
Kai Yu
This is my paper · ORCID
classification cs.CL
keywords toolhallucinationreliabilityalignmentlearningindecisiveactionspacepreferenceoptimizationRelyBenchReliablePassRateLLMagents
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 claims that a large share of LLM tool-use failures are avoidable hallucinations, and that models can be trained out of them by letting the model decline to call a tool when preconditions are unmet. It defines two hallucination families—selection (wrong or mistimed tool) and usage (wrong format or fabricated parameter content)—and builds RelyToolBench, a benchmark with missing-parameter and unmatched-tool cases, on top of StableToolBench. Its Relign method adds 'change tool' and 'talk to user' to the action space, trains with supervised fine-tuning plus direct preference optimization, and reports that on LLaMA-3.1-8B the tool hallucination rate drops from 50.8% to 14.6% while the reliable pass rate rises from 65.3% to 77.2%. The payoff, if true, is that reliability can be trained into tool-using LLMs at modest data scale rather than engineered around after the fact.

What carries the argument

The load-bearing mechanism is the indecisive action space: two deliberately non-decisive actions, ChangeTools and TalkToUser, that the model may output when tool preconditions are unmet. The paper also introduces RePR, the Reliable Pass Rate, defined as pass rate minus task hallucination rate, which discounts tasks whose final answers are contaminated by hallucinated tool calls. The preference hierarchy used to build DPO pairs—correct call preferred over indecisive action preferred over hallucinated call—is what turns the expanded space into a training signal.

What would settle it

Have human annotators independently label a random sample of RelyToolBench trajectories, including the final-answer correlation step, for the LLaMA-3.1 baseline and LLaMA-3.1 plus Relign; if the human labels do not reproduce the reported 50.8% to 14.6% hallucination drop and 65.3 to 77.2 RePR gain within reasonable error bars, the claim that Relign reduces hallucinations is not established.

Watch

Extended reading notes

Core claim

The paper's central claim is that tool hallucination is better modeled as a decision under missing preconditions than as a generic text-generation error, and that the remedy is to enlarge the model's action space so that abstention is a legitimate move. In Relign the model may still call the tool correctly, but may also invoke ChangeTools or TalkToUser, and the alignment objective orders trajectories as success > failure > hallucination, and actions as correct call > indecisive action > hallucinated call. Using SFT to teach the two new actions and DPO with preference pairs synthesized by a hallucination detector, the trained models reduce hallucination rates and tool-call counts across RelyToolBench, and the improvement transfers to APIBench without retraining on that benchmark.

Load-bearing premise

Everything hinges on GPT-4o judging hallucinations correctly, both when labeling training data and when scoring results, and the human check only covers 150 cases and not the step that decides whether a hallucinated call also corrupted the final answer.

Editorial extensions

If this is right

  • A model trained with Relign on LLaMA-3.1-8B cuts tool hallucination from 50.8% to 14.6% overall, and the reliable pass rate rises from 65.3% to 77.2%.
  • Relign lowers average tool calls per task (for example, LLaMA-3.1 goes from 2.2 to 1.5), so reliability gains also reduce compute and latency.
  • RePR is consistently below the original pass rate, showing that ordinary task success metrics overstate how often a task was completed without hallucination-tainted steps.
  • The gains transfer out of distribution: on APIBench, Relign reduces hallucination and improves tool-calling accuracy across HuggingFace, TorchHub, and TensorHub retrievals.
  • The paper's scaling analysis finds that larger models reduce hallucination under fixed data, while simply adding more ToolBench data does not, because that data lacks failure-case examples.

Reading between the lines

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

  • Inference: the abstention hierarchy is not tool-specific; any sequential decision task with detectable missing preconditions, such as web browsing, database queries, or robot commands, could use the same correct > indecisive > hallucinated preference ordering with domain-appropriate fallback actions.
  • Inference: because GPT-4o labels the preference pairs and scores the final metrics, the reported improvement could partly reflect overfitting to one judge; the paper's human validation covers only 150 cases and not the answer-correlation step that RePR relies on.
  • Inference: a natural testable extension is to re-run Relign's DPO pipeline with an open-weight judge or human preference labels; if the hallucination drop persists, the method is robust to judge choice, and if not, the measured gains are partially judge-fitting.
  • Inference: RePR treats hallucinated-call-to-answer correlation as a binary discount; future work could relax this to partial credit and estimate error bars on the discounted pass rate.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper addresses tool hallucination in LLM tool-use by (1) proposing a taxonomy (tool selection vs. tool usage, with four subtypes), (2) introducing RelyToolBench, a benchmark with missing-parameter and unmatched-tools subsets, along with two new metrics—Reliable Pass Rate (RePR) and Benefit-Cost Utility, and (3) presenting Relign, a reliability alignment framework that augments the tool-use action space with indecisive actions (ChangeTools, TalkToUser) and trains via SFT and DPO. Experiments on ToolLLaMA, LLaMA-3.1-8B, and Qwen2.5-7B report large reductions in tool hallucination rate (e.g., LLaMA-3.1: 50.8% to 14.6%) and improvements in RePR (65.3% to 77.2%), plus an OOD evaluation on APIBench.

Significance. If the results hold, the paper makes a useful contribution: it provides a systematic taxonomy of tool hallucinations, a benchmark that stresses failure modes not covered by existing tool benchmarks, and a practical training approach that yields substantial improvements across multiple base models. The release of code and data, the multi-model consistency, and the OOD generalization experiment are concrete strengths. The significance is tempered, however, by the fact that the headline claims rest on a single LLM evaluator (GPT-4o) that is also used to construct the DPO training signal, and the human validation of that evaluator is narrow. The metric definitions also need correction. These issues are addressable but currently leave the central quantitative claims less certain than the paper suggests.

major comments (3)
  1. [§3.2.2, §4.1, Appendix A] The same GPT-4o model is used both to label hallucination types for DPO preference construction (§3.2.2) and to compute the reported tool hallucination rate and RePR (§4.1). The human validation in Appendix A covers only 150 cases across three categories (no hallucination, parameter value hallucination, tool relevance hallucination) and does not validate timing or format hallucinations, nor does it validate the Appendix D answer-correlation step that converts tool hallucinations into the task hallucination rate subtracted in RePR. Since RePR and the hallucination-rate reductions are the paper's central claims, this evaluator circularity is load-bearing. I recommend validating the evaluator with a second independent LLM or a larger human sample covering all four subtypes and the answer-correlation step, and reporting agreement statistics separately for each component.
  2. [§2.2.2, Eq. (Utility)] The definition of P_hallucination is internally inconsistent. The text states 'P_hallucination is the penalty for hallucinations in the task (-10 for hallucination)' but the formula subtracts P_hallucination. If P_hallucination is negative, subtracting it adds to the utility, which would reward hallucinations; if the intended penalty is a nonnegative magnitude of 10, the parenthetical is misleading. Please clarify the sign convention and recompute the Utility values in Table 2 under a consistent definition, or explain why the current convention is correct.
  3. [§4.2, Table 2 caption] The table caption reads 'Utility: Benifit-cost utility↓' while the column header and the text indicate that higher Utility is better (GPT-4o scores 13.4 vs. 5.2 for GPT-3.5-turbo, and the text describes utility as a measure to be maximized). This direct contradiction in the direction of the arrow makes the utility results ambiguous. Please correct the caption and ensure the arrow direction matches the text and the reported numbers.
minor comments (6)
  1. [§2.2.2] The word 'Benifit' is misspelled; it should be 'Benefit'.
  2. [§3.2.2] The term 'Acoreect' appears to be a typo for 'Acorrect'.
  3. [§4.1] The phrase 'For computing both benefit-cost utility' is ungrammatical; it should read 'For computing the benefit-cost utility'.
  4. [§3.2.1] The sentence 'the training data we selected involves only a single tool' is ambiguous—does it mean each training example contains only one tool call, or that the examples are restricted to a single-tool task? Please clarify.
  5. [Appendix A] The human evaluation is described as validating 'outputs generated by GPT-4, GPT-3.5, and ToolLLaMA3.1,' but the paper's evaluator is GPT-4o; please specify which model's outputs were judged and whether the human agreement applies directly to the GPT-4o-based pipeline.
  6. [Table 2] The column header 'Utility↑' is inconsistent with the caption's 'Benifit-cost utility↓' (see major comment); beyond the direction, consider using a consistent notation such as 'Utility ↑' and 'Tool Hallu ↓' in both the header and caption.

Circularity Check

1 steps flagged · score 4.0 of 10

The headline hallucination reduction is measured by the same GPT-4o judge used to build the DPO preference data, and the human check covers only 150 cases without validating the answer-correlation step inside RePR.

  1. fitted input called prediction [Section 3.2.2 (Reliable DPO) and Section 4.1 (Experimental Setup); Table 2]
    "In each round, we sample ten tool-use trajectories at a temperature of 0.7 and use GPT-4O as a hallucination evaluator to construct preference pairs. ... Additionally, evaluations regarding tool hallucinations and task success were performed using the GPT-4 O model"

    The DPO training signal is generated by GPT-4o classifying tool calls as hallucinated, non-hallucinated, or indecisive, and the same GPT-4o judge computes the reported tool hallucination rate and RePR components in Table 2. Thus the headline reductions (e.g., LLaMA-3.1-8B Hallu dropping from 50.8 to 14.6 and RePR rising from 65.3 to 77.2) partly measure how well Relign matches the biases of the judge that created its preference labels, rather than an independent property. The Appendix A human validation covers only 150 cases across three categories and does not validate the Appendix D answer-correlation step that converts tool hallucinations into Task Hallucination Rate inside RePR, so the load-bearing part of the metric depends on the same unvalidated judge used for training.

full rationale

This paper does not present a mathematical derivation, so the circularity question is about the evaluation loop. The central empirical claim is that Relign reduces tool hallucination and improves reliability, but the judge used to label the DPO preference data (Section 3.2.2: 'use GPT-4O as a hallucination evaluator to construct preference pairs') is the same model used to compute the reported hallucination and task-reliability metrics (Section 4.1: 'evaluations regarding tool hallucinations and task success were performed using the GPT-4 O model'). Any systematic bias in the GPT-4o judge can be learned by the model during preference optimization and then rewarded by the same judge during evaluation, so the reported reduction is partially a fit to the judge rather than an independently verified reduction in hallucination. The human evaluation in Appendix A is small (150 cases) and covers only no hallucination, parameter value hallucination, and tool relevance hallucination; it does not validate timing or format hallucination judgments, and it does not validate the Appendix D answer-correlation step used by RePR. Multi-model consistency and public code are real supporting evidence, but they do not break the dependence on a single judge. The score is 4 rather than higher because the method also improves on external measures such as APIBench AST accuracy and the pass-rate component of RePR, so the central claim is not entirely determined by the judge alone.

Assumptions & free parameters 4 free parameters · 4 assumptions · 1 invented entities

The central claim depends on several hand-set constants in the utility metric, on normative preference hierarchies, and on the assumption that GPT-4o is a reliable judge. The free parameters mostly affect the reported utility values rather than the pass-rate or hallucination-rate findings, but the judge assumption is load-bearing because it was used both to create training labels and to measure outcomes.

free parameters (4)
  • R_task success reward = 20
    Hand-chosen constant in the Benefit-cost Utility metric; the value affects utility magnitudes but not the ranking of methods in Table 2 (no sensitivity analysis is provided).
  • P_hallucination penalty = -10 (as written)
    Hand-chosen penalty for hallucinated tasks; the sign is ambiguous in the text ('(-10 for hallucination)') and appears inconsistent with the formula that subtracts it.
  • P_tool penalty cap = 10
    Caps the penalty for excessive tool calls in the utility metric; no principled justification is given.
  • necessary tool call count = 1 for solvable, 0 for unsolvable
    Per-task assumption used in utility computation; not derived from data and may penalize models that legitimately need multiple calls.
assumptions (4)
  • domain assumption Preference hierarchy Tsuccess > Tfailure > Thallucination
    Stated in Section 3.1; assumes a failed task is always preferable to a hallucinated result, which is a safety-oriented normative assumption not empirically validated.
  • domain assumption Step-level preference Acorrect > Aindecisive > Ahallucinated
    Section 3.1; assumes indecisive actions are better than hallucinated calls, which underpins the DPO data construction.
  • domain assumption The four-subtype taxonomy of tool hallucinations is exhaustive and meaningful
    Section 2.1 defines two types and four subtypes; whether this categorization covers all failure modes is not empirically grounded.
  • domain assumption GPT-4o-based evaluator agrees with human judgment in the full evaluation setup
    Section 2.1 and Appendix A; human evaluation covers only 150 cases and three categories, not all subtypes or the answer-correlation step of RePR.
invented entities (1)
  • Indecisive action space (ChangeTools, TalkToUser)
    purpose: Allows the model to defer tool invocation, ask clarifying questions, or switch tools when preconditions are unmet.
    These actions are introduced by the paper's alignment framework; they are evaluated only within the paper's own benchmark and have no external falsifiable handle beyond the reported experiments.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Reducing Tool Hallucination via Reliability Alignment." pith.science (2026). https://pith.science/paper/TV5X63E6

@misc{pith2026241204141,
  author       = {Pith},
  title        = {Pith review of: Reducing Tool Hallucination via Reliability Alignment},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TV5X63E6}},
  note         = {Machine review of arXiv:2412.04141}
}
read the original abstract

Large Language Models (LLMs) have expanded their capabilities beyond language generation to interact with external tools, enabling automation and real-world applications. However, tool hallucinations, where models either select inappropriate tools or misuse them, pose significant challenges, leading to erroneous task execution, increased computational costs, and reduced system reliability. To systematically address this issue, we define and categorize tool hallucinations into two main types, tool selection hallucination and tool usage hallucination. To evaluate and mitigate these issues, we introduce RelyToolBench, which integrates specialized test cases and novel metrics to assess hallucination-aware task success and efficiency. Finally, we propose Relign, a reliability alignment framework that expands the tool-use action space to include indecisive actions, allowing LLMs to defer tool use, seek clarification, or adjust tool selection dynamically. Through extensive experiments, we demonstrate that Relign significantly reduces tool hallucinations, improves task reliability, and enhances the efficiency of LLM tool interactions.

Figures

Figures reproduced from arXiv: 2412.04141 by the authors.

Figure 1
Figure 1. Different types of tool hallucination. With the emergence of intelligent agent frameworks like AutoGPT and MetaGPT (Hong et al., 2024), LLMs now demonstrate enhanced capabilities in real-world interactions through tool use. However, when LLMs interface with external environments via these tools, hallucinations during tool execution can directly affect real-world systems. We view tool calling as a decision-making act… view at source ↗
Figure 2
Figure 2. Evaluation process of tool hallucination. • Tool type hallucination indicates that the model either invokes a tool that is unrelated to the task or fabricates a tool that does not exist in the available tool set. • Tool timing hallucination involves errors related to the sequence of tool calling. Specifically, this form of hal￾lucination occurs when the model calls the same tool repeatedly with identical inputs and … view at source ↗
Figure 3
Figure 3. Metric comparison between reliable and original pass rate. O, MP, and UT represent the original, missing parameter, and unmatched tools subsets, respectively. 2.2.2. BENEFIT-COST UTILITY The Benefit-Cost Utility takes a more holistic approach by evaluating both the task outcome and tool usage efficiency. The utility score is designed to reflect the overall quality of task execution, considering not only task success… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: The system illustration of Relign. 3.1. Alignment Goal The alignment goal for reliable tool calling in LLMs focuses on ensuring accurate and effective interactions with external tools throughout the task completion process. From a task perspective, the primary objectiv…
Figure 5
Figure 5. Figure 5: Comparison of performance metrics between the baseline and Relign across three subsets: Original (O), Missing Parameter (MP), and Unmatched Tools (UT). duction in tool hallucination rates and do not decrease the number of tool calls. In contrast, our approach not only …
Figure 7
Figure 7. Figure 7: Impact of Training Data Size on Performance and Hallucination Metrics. Experiments are conducted on LLaMA3.1. 1.5B 3B 7B 14B Model Size 20 30 40 50 60 70 Reliable Pass Rate 1.5B 3B 7B 14B Model Size 30 40 50 60 70 80 90 Tool Hallucination Rate 0 3 6 9 12 15 Avg Tool Ca…
Figure 8
Figure 8. Figure 8: Impact of Model Size on Performance and Halluci￾nation Metrics. Experiments are conducted on Qwen2.5 with 5k data. that reducing tool hallucinations helps models learn better tool-use strategies (AST represents tool-calling accuracy). Method HuggingFace TorchHub Tensor…
Figure 9
Figure 9. Figure 9: A case study on the missing parameter subset comparing the Baseline method and Relign: Relign successfully corrects the parameter hallucination. 15 [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 4 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. AgentOPSD: Recursive Self-Distillation for Agentic Reinforcement Learning

    cs.AI 2026-08 conditional novelty 7.0 of 10

    AgentOPSD converts sparse outcome rewards into turn-level credit by recursively accumulating self-distillation evidence in log-odds space, beating GRPO on most agentic benchmarks.

  2. PredAct-Bench: Benchmarking Tool-Augmented Dialogue under Controlled Tool Noise

    cs.CL 2026-08 conditional novelty 6.0 of 10

    When grade-prediction tools are noisy, most LLM instructors over-rely on them in multi-turn dialogue and their decisions degrade, whereas human instructors stay better calibrated.

  3. Multi-Head Latent Control: A Unified Interface for LLM Agent Decision Making

    cs.CL 2026-07 conditional novelty 6.0 of 10

    Hidden-state traces of frozen LLMs/VLMs can be read by lightweight trained heads to predict when to defer, clarify, call tools, or abstain, cutting routed inference cost 27–90%.

  4. Enhancing Tool Learning in Large Language Models with Hierarchical Error Checklists

    cs.CL 2025-05 conditional novelty 6.0 of 10

    HiTEC improves LLM tool calling by embedding hierarchical error checklists in prompts or using them to generate negative examples for KTO fine-tuning.

Reference graph

Works this paper leans on

44 extracted references · 8 canonical work pages · cited by 4 Pith papers

  1. [1]

    L., Almeida, D., Altenschmidt, J., Altman, S., Anadkat, S., et al

    Achiam, J., Adler, S., Agarwal, S., Ahmad, L., Akkaya, I., Aleman, F. L., Almeida, D., Altenschmidt, J., Altman, S., Anadkat, S., et al. Gpt-4 technical report. arXiv preprint arXiv:2303.08774, 2023

    arXiv 2023

  2. [2]

    and Mitchell, T

    Azaria, A. and Mitchell, T. The internal state of an llm knows when its lying. arXiv preprint arXiv:2304.13734, 2023

    arXiv 2023

  3. [3]

    Large Language Models as Tool Makers

    Cai, T., Wang, X., Ma, T., Chen, X., and Zhou, D. Large Language Models as Tool Makers . In Proc. of The Twelfth International Conference on Learning Representations (ICLR 2024) , 2024. URL https://openreview.net/forum?id=qV83K9d5WB

    work page 2024

  4. [4]

    Alpagasus: Training a better alpaca with fewer data

    Chen, L., Li, S., Yan, J., Wang, H., Gunaratna, K., Yadav, V., Tang, Z., Srinivasan, V., Zhou, T., Huang, H., et al. Alpagasus: Training a better alpaca with fewer data. arXiv preprint arXiv:2307.08701, 2023 a

    arXiv 2023

  5. [5]

    T-Eval: Evaluating the Tool Utilization Capability Step by Step

    Chen, Z., Du, W., Zhang, W., Liu, K., Liu, J., Zheng, M., Zhuo, J., Zhang, S., Lin, D., Chen, K., et al. T-Eval: Evaluating the Tool Utilization Capability Step by Step . ArXiv preprint, abs/2312.14033, 2023 b . URL https://arxiv.org/abs/2312.14033

    arXiv 2023

  6. [6]

    The llama 3 herd of models

    Dubey, A., Jauhri, A., Pandey, A., Kadian, A., Al-Dahle, A., Letman, A., Mathur, A., Schelten, A., Yang, A., Fan, A., et al. The llama 3 herd of models. arXiv preprint arXiv:2407.21783, 2024

    arXiv 2024

  7. [7]

    Retrieval-augmented generation for large language models: A survey

    Gao, Y., Xiong, Y., Gao, X., Jia, K., Pan, J., Bi, Y., Dai, Y., Sun, J., and Wang, H. Retrieval-augmented generation for large language models: A survey. arXiv preprint arXiv:2312.10997, 2023

    arXiv 2023

  8. [8]

    Gemini: A Family of Highly Capable Multimodal Models , 2023

    Gemini Team . Gemini: A Family of Highly Capable Multimodal Models , 2023

    work page 2023

Show all 44 references
  1. [9]

    M., Alves, D

    Guerreiro, N. M., Alves, D. M., Waldendorf, J., Haddow, B., Birch, A., Colombo, P., and Martins, A. F. Hallucinations in large multilingual translation models. Transactions of the Association for Computational Linguistics, 11: 0 1500--1517, 2023

    2023

  2. [10]

    Stabletoolbench: Towards stable large-scale benchmarking on tool learning of large language models

    Guo, Z., Cheng, S., Wang, H., Liang, S., Qin, Y., Li, P., Liu, Z., Sun, M., and Liu, Y. Stabletoolbench: Towards stable large-scale benchmarking on tool learning of large language models. arXiv preprint arXiv:2403.07714, 2024

    2024 arXiv

  3. [11]

    and Kembhavi, A

    Gupta, T. and Kembhavi, A. Visual programming: Compositional visual reasoning without training. In In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2023) , pp.\ 14953--14962, 2023

    2023

  4. [12]

    ToolkenGPT: Augmenting Frozen Language Models with Massive Tools via Tool Embeddings

    Hao, S., Liu, T., Wang, Z., and Hu, Z. ToolkenGPT: Augmenting Frozen Language Models with Massive Tools via Tool Embeddings . ArXiv preprint, abs/2305.11554, 2023. URL https://arxiv.org/abs/2305.11554

    2023 arXiv

  5. [13]

    Hong, S., Zhuge, M., Chen, J., Zheng, X., Cheng, Y., Wang, J., Zhang, C., Wang, Z., Yau, S. K. S., Lin, Z., Zhou, L., Ran, C., Xiao, L., Wu, C., and Schmidhuber, J. Meta GPT : Meta programming for a multi-agent collaborative framework. In The Twelfth International Conference o...

    2024

  6. [14]

    Tool documentation enables zero-shot tool-usage with large language models

    Hsieh, C.-Y., Chen, S.-A., Li, C.-L., Fujii, Y., Ratner, A., Lee, C.-Y., Krishna, R., and Pfister, T. Tool documentation enables zero-shot tool-usage with large language models. ArXiv preprint, abs/2308.00675, 2023. URL https://arxiv.org/abs/2308.00675

    2023 arXiv

  7. [15]

    K., and Kumar, V

    Jain, N., Kwiatkowski, R., Ray, B., Ramanathan, M. K., and Kumar, V. On mitigating code llm hallucinations with api documentation. arXiv preprint arXiv:2407.09726, 2024

    2024 arXiv

  8. [16]

    J., Madotto, A., and Fung, P

    Ji, Z., Lee, N., Frieske, R., Yu, T., Su, D., Xu, Y., Ishii, E., Bang, Y. J., Madotto, A., and Fung, P. Survey of hallucination in natural language generation. ACM Computing Surveys, 55 0 (12): 0 1--38, 2023

    2023

  9. [17]

    GeneGPT: augmenting large language models with domain tools for improved access to biomedical information

    Jin, Q., Yang, Y., Chen, Q., and Lu, Z. GeneGPT: augmenting large language models with domain tools for improved access to biomedical information . Bioinformatics, 40 0 (2): 0 btae075, 2024. ISSN 1367-4811. doi:10.1093/bioinformatics/btae075. URL https://doi.org/10.1093/bioinf...

    2024 doi

  10. [18]

    Crafting papers on machine learning

    Langley, P. Crafting papers on machine learning. In Langley, P. (ed.), Proceedings of the Seventeenth International Conference on Machine Learning (ICML 2000), Stanford University, Stanford, CA, USA, June 29 - July 2, 2000 , pp.\ 1207--1216. Morgan Kaufmann, 2000

    2000

  11. [19]

    W., Iyyer, M., Zettlemoyer, L., and Hajishirzi, H

    Min, S., Krishna, K., Lyu, X., Lewis, M., Yih, W.-t., Koh, P. W., Iyyer, M., Zettlemoyer, L., and Hajishirzi, H. Factscore: Fine-grained atomic evaluation of factual precision in long form text generation. arXiv preprint arXiv:2305.14251, 2023

    2023 arXiv

  12. [20]

    Self-contradictory hallucinations of large language models: Evaluation, detection and mitigation

    M \"u ndler, N., He, J., Jenko, S., and Vechev, M. Self-contradictory hallucinations of large language models: Evaluation, detection and mitigation. arXiv preprint arXiv:2305.15852, 2023

    2023 arXiv

  13. [21]

    G., Zhang, T., Wang, X., and Gonzalez, J

    Patil, S. G., Zhang, T., Wang, X., and Gonzalez, J. E. Gorilla: Large Language Model Connected with Massive APIs . ArXiv preprint, abs/2305.15334, 2023. URL https://arxiv.org/abs/2305.15334

    2023 arXiv

  14. [22]

    The RefinedWeb dataset for Falcon LLM: outperforming curated corpora with web data, and web data only

    Penedo, G., Malartic, Q., Hesslow, D., Cojocaru, R., Cappelli, A., Alobeidli, H., Pannier, B., Almazrouei, E., and Launay, J. The RefinedWeb dataset for Falcon LLM: outperforming curated corpora with web data, and web data only . ArXiv preprint, abs/2306.01116, 2023. URL https...

    2023 arXiv

  15. [23]

    WebCPM: Interactive Web Search for Chinese Long-form Question Answering

    Qin, Y., Cai, Z., Jin, D., Yan, L., Liang, S., Zhu, K., Lin, Y., Han, X., Ding, N., Wang, H., et al. WebCPM: Interactive Web Search for Chinese Long-form Question Answering . ArXiv preprint, abs/2305.06849, 2023 a . URL https://arxiv.org/abs/2305.06849

    2023 arXiv

  16. [24]

    ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs , 2023 b

    Qin, Y., Liang, S., Ye, Y., Zhu, K., Yan, L., Lu, Y., Lin, Y., Cong, X., Tang, X., Qian, B., Zhao, S., Tian, R., Xie, R., Zhou, J., Gerstein, M., Li, D., Liu, Z., and Sun, M. ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs , 2023 b

    2023

  17. [25]

    Qin, Y., Hu, S., Lin, Y., Chen, W., Ding, N., Cui, G., Zeng, Z., Huang, Y., Xiao, C., Han, C., Fung, Y. R., Su, Y., Wang, H., Qian, C., Tian, R., Zhu, K., Liang, S., Shen, X., Xu, B., Zhang, Z., Ye, Y., Li, B., Tang, Z., Yi, J., Zhu, Y., Dai, Z., Yan, L., Cong, X., Lu, Y., Zha...

    2024 arXiv

  18. [26]

    D., Ermon, S., and Finn, C

    Rafailov, R., Sharma, A., Mitchell, E., Manning, C. D., Ermon, S., and Finn, C. Direct preference optimization: Your language model is secretly a reward model. Advances in Neural Information Processing Systems, 36, 2024

    2024

  19. [27]

    Toolformer: Language Models Can Teach Themselves to Use Tools

    Schick, T., Dwivedi-Yu, J., Dess \` , R., Raileanu, R., Lomeli, M., Zettlemoyer, L., Cancedda, N., and Scialom, T. Toolformer: Language Models Can Teach Themselves to Use Tools . ArXiv preprint, abs/2302.04761, 2023. URL https://arxiv.org/abs/2302.04761

    2023 arXiv

  20. [28]

    Shi, W., Han, X., Lewis, M., Tsvetkov, Y., Zettlemoyer, L., and Yih, S. W.-t. Trusting your evidence: Hallucinate less with context-aware decoding. arXiv preprint arXiv:2305.14739, 2023

    2023 arXiv

  21. [29]

    ToolAlpaca: Generalized Tool Learning for Language Models with 3000 Simulated Cases , 2023

    Tang, Q., Deng, Z., Lin, H., Han, X., Liang, Q., and Sun, L. ToolAlpaca: Generalized Tool Learning for Language Models with 3000 Simulated Cases , 2023

    2023

  22. [30]

    Llama 2: Open foundation and fine-tuned chat models

    Touvron, H., Martin, L., Stone, K., Albert, P., Almahairi, A., Babaei, Y., Bashlykov, N., Batra, S., Bhargava, P., Bhosale, S., et al. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288, 2023

    2023 arXiv

  23. [31]

    A stitch in time saves nine: Detecting and mitigating hallucinations of llms by validating low-confidence generation

    Varshney, N., Yao, W., Zhang, H., Chen, J., and Yu, D. A stitch in time saves nine: Detecting and mitigating hallucinations of llms by validating low-confidence generation. arXiv preprint arXiv:2307.03987, 2023

    2023 arXiv

  24. [32]

    Can llms express their uncertainty? an empirical evaluation of confidence elicitation in llms

    Xiong, M., Hu, Z., Lu, X., Li, Y., Fu, J., He, J., and Hooi, B. Can llms express their uncertainty? an empirical evaluation of confidence elicitation in llms. arXiv preprint arXiv:2306.13063, 2023

    2023 arXiv

  25. [33]

    Rejection improves reliability: Training llms to refuse unknown questions using rl from knowledge feedback

    Xu, H., Zhu, Z., Ma, D., Zhang, S., Fan, S., Chen, L., and Yu, K. Rejection improves reliability: Training llms to refuse unknown questions using rl from knowledge feedback. arXiv preprint arXiv:2403.18349, 2024

    2024 arXiv

  26. [34]

    Alignment for efficient tool calling of large language models

    Xu, H., Wang, Z., Zhu, Z., Pan, L., Chen, X., Chen, L., and Yu, K. Alignment for efficient tool calling of large language models. arXiv preprint arXiv:2503.06708, 2025

    2025 arXiv

  27. [35]

    Yang, A., Yang, B., Zhang, B., Hui, B., Zheng, B., Yu, B., Li, C., Liu, D., Huang, F., Wei, H., et al. Qwen2. 5 technical report. arXiv preprint arXiv:2412.15115, 2024

    2024 arXiv

  28. [36]

    ChatGPT is not Enough: Enhancing Large Language Models with Knowledge Graphs for Fact-aware Language Modeling

    Yang, L., Chen, H., Li, Z., Ding, X., and Wu, X. ChatGPT is not Enough: Enhancing Large Language Models with Knowledge Graphs for Fact-aware Language Modeling . ArXiv preprint, abs/2306.11489, 2023. URL https://arxiv.org/abs/2306.11489

    2023 arXiv

  29. [37]

    WebShop: Towards Scalable Real-World Web Interaction with Grounded Language Agents

    Yao, S., Chen, H., Yang, J., and Narasimhan, K. WebShop: Towards Scalable Real-World Web Interaction with Grounded Language Agents . In Koyejo, S., Mohamed, S., Agarwal, A., Belgrave, D., Cho, K., and Oh, A. (eds.), In Proceedings of the Advances in Neural Information Processi...

    2022

  30. [38]

    Steptool: A step-grained reinforcement learning framework for tool learning in llms

    Yu, Y., Wang, Z., Ma, W., Guo, Z., Zhan, J., Wang, S., Wu, C., Guo, Z., and Zhang, M. Steptool: A step-grained reinforcement learning framework for tool learning in llms. arXiv preprint arXiv:2410.07745, 2024

    2024 arXiv

  31. [39]

    Instruction tuning for large language models: A survey

    Zhang, S., Dong, L., Li, X., Zhang, S., Sun, X., Wang, S., Li, J., Hu, R., Zhang, T., Wu, F., et al. Instruction tuning for large language models: A survey. arXiv preprint arXiv:2308.10792, 2023

    2023

  32. [40]

    Toolbehonest: A multi-level hallucination diagnostic benchmark for tool-augmented large language models

    Zhang, Y., Chen, J., Wang, J., Liu, Y., Yang, C., Shi, C., Zhu, X., Lin, Z., Wan, H., Yang, Y., et al. Toolbehonest: A multi-level hallucination diagnostic benchmark for tool-augmented large language models. arXiv preprint arXiv:2406.20015, 2024

    2024 arXiv

  33. [41]

    Verify-and-edit: A knowledge-enhanced chain-of-thought framework

    Zhao, R., Li, X., Joty, S., Qin, C., and Bing, L. Verify-and-edit: A knowledge-enhanced chain-of-thought framework. arXiv preprint arXiv:2305.03268, 2023

    2023 arXiv

  34. [42]

    Enhancing llm reliability via explicit knowledge boundary modeling

    Zheng, H., Xu, H., Liu, Y., Chen, L., Fung, P., and Yu, K. Enhancing llm reliability via explicit knowledge boundary modeling. arXiv preprint arXiv:2503.02233, 2025

    2025

  35. [43]

    Lima: Less is more for alignment

    Zhou, C., Liu, P., Xu, P., Iyer, S., Sun, J., Mao, Y., Ma, X., Efrat, A., Yu, P., Yu, L., et al. Lima: Less is more for alignment. arXiv preprint arXiv:2305.11206, 2023

    2023 arXiv

  36. [44]

    ToolQA: A Dataset for LLM Question Answering with External Tools

    Zhuang, Y., Yu, Y., Wang, K., Sun, H., and Zhang, C. ToolQA: A Dataset for LLM Question Answering with External Tools . ArXiv preprint, abs/2306.13304, 2023. URL https://arxiv.org/abs/2306.13304

    2023 arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.