Pith. sign in

REVIEW 4 major objections 6 minor 41 references

Empowering LLMs in Task-Oriented Dialogues: A Domain-Independent Multi-Agent Framework and Fine-Tuning Strategy

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A fine-tuned 7B model organized as three domain-independent agents outperforms larger and traditional baselines on the MultiWOZ 2.2 task-oriented dialogue benchmark.

desk verdict A sensible task-level multi-agent decomposition for TOD that posts a strong MultiWOZ number, but the DDA contribution rests on a confounded comparison and the evaluation is single-run. read the letter →

arxiv 2505.14299 v1 pith:2RMAJEQH submitted 2025-05-20 cs.MA

Zihao Feng , Xiaoxue Wang , Bowen Wu , Weihong Zhong , Zhen Xu , Hailong Cao , Tiejun Zhao , Ying Li
show 1 more author
Baoxun Wang
This is my paper
classification cs.MA MSC 68T5068T42
keywords task-orienteddialoguemulti-agentframeworkdomain-independentagentsdirectpreferenceoptimizationdatadistributionadaptationMultiWOZ2.2zero-shotgeneralizationlightweightLLMs
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to show that a relatively small open LLM can run a task-oriented dialogue system at a level that beats larger, more expensive systems, provided the job is split into three simple, domain-independent agents instead of one monolithic agent. To train those agents, it applies direct preference optimization after supervised fine-tuning, and it adds a data-balancing procedure called Data Distribution Adaptation that keeps DPO from degrading on domains underrepresented among rejected examples. On MultiWOZ 2.2, the system reaches a Combined score of 106.3, the best average among all compared methods, including a replicated multi-agent baseline on the same 7B base and a larger proprietary model. If the claim holds, competitive task-oriented dialogue can come from a fine-tuned 7B model with a modular framework, not only from huge closed models.

What carries the argument

The machinery is two stacked mechanisms. The Domain-Independent Multi-Agent Framework (DIMF) splits a dialogue into three agents—Intent Classification, Slot Filling, and Response—each with its own prompt containing logical rules for multi-turn behavior, so a single 7B model can handle all domains instead of training separate domain-specific agents. The Data Distribution Adaptation (DDA) method generates synthetic bad cases for underrepresented domains and slot categories so that the rejected examples in DPO mirror the domain distribution of the SFT data, counteracting DPO's tendency to suppress whatever category dominates the rejected set; the paper shows reward curves where DPO-DDA keeps the chosen reward positive while original DPO's chosen reward drops below zero.

What would settle it

Train the Slot Filling Agent with the original bad cases plus the same number of synthetic bad cases concentrated in one extra domain (no balancing) and measure the Combined score; if it matches DPO-DDA's improvement, the effect is data quantity, not distribution adaptation.

Watch

Extended reading notes

Core claim

The paper claims that decomposing a task-oriented dialogue system into three domain-independent agents—Intent Classification, Slot Filling, and Response—lets a lightweight fine-tuned LLM (Qwen2.5-7B) outperform both traditional fine-tuned systems and larger LLM-based systems on MultiWOZ 2.2. It further claims that the Data Distribution Adaptation method fixes a specific failure of DPO: when rejected examples concentrate unevenly across domains, DPO suppresses the model's output for those categories, and adding synthetic bad examples to rebalance the rejected data restores stable training. The paper's own framing is that the domain-independent decomposition reduces each agent's learning complexity, while DDA-DPO sharpens the model's ability to follow the logical rules in the prompts without the usual DPO degradation.

Load-bearing premise

The load-bearing premise is that the improvement from the data-balancing trick comes specifically from balancing the kinds of rejected examples across domains, not from simply adding more training data; no experiment in the paper separates those two effects.

Editorial extensions

If this is right

  • A fine-tuned 7B open-weight model can beat much larger LLM-based dialogue systems on MultiWOZ 2.2 when the workflow is decomposed into domain-independent agents, reaching Combined 106.3 versus 101.3 for the Claude Sonnet 3.0 DARD baseline.
  • DPO can be safely applied to intermediate dialogue agents if the rejected data is rebalanced: DDA improves Inform and Success over original DPO on all five test domains for the Slot Filling Agent.
  • Removing one domain's training data leaves the system's overall score largely unchanged for Attraction, Restaurant, and Taxi, indicating zero-shot transfer across domains.
  • Ablations show that each additional agent separation (single-agent to two-agent to three-agent SFT) raises Inform and Success, and that each DPO stage on top of SFT reduces error accumulation.

Reading between the lines

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

  • As an extension, the DDA balancing principle is not specific to dialogue: any DPO pipeline whose rejected examples cluster in one output category could be stabilized by synthesizing rejected examples for underrepresented categories, a claim the paper does not test.
  • Because DIMF uses one model with per-domain prompts, the same domain-independent decomposition could be lifted onto other task-oriented benchmarks or real API collections by swapping the prompt's tool descriptions, giving a direct test of the claimed generalization.
  • The paper's own limitation note says the database lookup is rule-based; porting DIMF to real tools would require the model to generate a unified invocation statement, so the framework's current gains may not carry over to tool-rich environments without additional machinery.
  • The paper names GRPO as future work; a plausible next step is to combine DDA-style distribution balancing with online preference optimization, since DDA addresses only offline DPO's rejected-data skew.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper addresses task-oriented dialogue with lightweight LLMs by decomposing the pipeline into three domain-independent agents (intent classification, slot filling, response), fine-tuning Qwen2.5-7B with SFT followed by DPO, and introducing a Data Distribution Adaptation (DDA) method that generates synthetic bad cases to balance the domain distribution of rejected preference pairs. On MultiWOZ 2.2, the full system reports Combined 106.3 (Inform 92.4, Success 82.8, BLEU 18.7), outperforming the cited baselines including SUIT2 (105.1) and Claude Sonnet 3.0 DARD (101.3). The paper also reports domain-wise DPO comparisons, zero-shot leave-one-domain-out experiments, framework ablations, and a case study.

Significance. The central result is practically interesting: if reproducible, a fine-tuned 7B model can exceed much larger LLM agents on a standard task-oriented dialogue benchmark. The DIMF decomposition is simple and the DDA mechanism is a plausible remedy for DPO degradation. The paper uses the standard MultiWOZ 2.2 evaluation and reports diversity metrics in addition to the standard Inform/Success/BLEU. However, the significance is currently limited because the headline SOTA rests on a single run without variance, the DDA effect is confounded with data quantity, the zero-shot experiment is weakened by the presence of domain information in prompts, and the DARD baseline replication is not documented. These are fixable with additional controls and reporting.

major comments (4)
  1. [§6.2, Appendix B] The comparison between DPO-Ori and DPO-DDA in Table 2 varies two factors simultaneously: the domain distribution of rejected examples and the total number of preference pairs. Appendix B describes DDA as adding synthetic bad cases to the original bad cases, so DPO-DDA always has more preference pairs than DPO-Ori. The paper's causal claim that degradation is caused by imbalanced rejected data and that DDA mitigates it by matching the SFT distribution requires a control that holds the number of pairs fixed, e.g., adding the same number of extra bad cases sampled from the original imbalanced distribution. Without this DPO-Quantity control, the improvements in Table 2 and Figure 3 may be explained by data quantity rather than distribution adaptation. This concern is load-bearing for the paper's second contribution.
  2. [Table 1, §6.1] All reported results appear to be single runs with no error bars or significance tests. The Combined-score advantage of the proposed method over the closest baseline SUIT2 is 1.2 points (106.3 vs. 105.1), which is small relative to typical fine-tuning variance. Please provide multiple random seeds and confidence intervals, at least for Table 1 and the DPO-DDA vs. DPO-Ori comparisons in Table 2, to support the SOTA claim. Without this, the 'better average performance' claim is not statistically grounded.
  3. [§6.3, Appendix A] Section 6.3 removes a domain's training data but keeps the domain's information in the prompts: the Appendix A prompts list all domain tools, slots, and allowed values (e.g., find_train in Table 5 and restaurant slots in Table 6). The zero-shot experiment therefore tests the model's ability to operate without training examples for a domain whose schema is fully specified at inference, not its ability to generalize to a genuinely unseen domain. The contribution statement that the system maintains good performance 'even on domains it has not seen before' is stronger than what the experiment establishes. Either hold out the prompt schema as well, or revise the zero-shot claim to 'no training examples for the domain'.
  4. [§5.2, Table 1] Section 5.2 states that DARD code was not provided and that the authors independently replicated it, but no replication details are given (prompts, inference procedure, hyperparameters, or number of runs). The reader cannot assess whether the DARD baseline was given a fair or strong configuration. Please provide the exact replication protocol, or report the comparison without the unverifiable DARD numbers in the headline table.
minor comments (6)
  1. [Figures 3 and 7] The captions of Figures 3 and 7 use 'DPO-DDT' while the text and method name are 'DDA'; unify the notation.
  2. [Table 7] Table 7 contains typos ('repone', 'Waht's', 'Responce Rules') and ungrammatical phrasing ('reponse user's question relied some given informations'); please proofread the appendix prompts.
  3. [Table 1 footnote] The footnote says bold indicates the best score 'among all the systems of each language pair', but the table is not split by language pair; clarify the intended grouping.
  4. [§6.2] The degradation criterion 'drops below the average accuracy' is ad hoc; state the threshold and justify it, or compare per-domain scores against the SFT baseline instead.
  5. [§5.2, Appendix B] Section 5.2 promises training settings in Appendix B, but Appendix B only describes DDA data generation; include the actual hyperparameters (learning rate, batch size, DPO beta, number of DDA-generated cases per domain, BLEU threshold for Response Agent positives) or correct the cross-reference.
  6. [References] The reference to Gupta et al. (DARD) lacks a full citation with year and venue; complete the reference.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation found; headline evaluation is held-out and the DDA comparison has a validity confound, not a circularity.

full rationale

The headline result is self-contained: Table 1 evaluates on the held-out MultiWOZ 2.2 test split, and the DPO preference pairs are explicitly taken from the validation set (Figure 2 caption: 'We first fine-tune the model with the training set, and then leverage the validation dataset to complete the DPO process'), so the reported Combined score of 106.3 is not produced by fitting to the test set or by a parameter that is defined as the target metric. The DDA analysis in Section 6.2 and Appendix B compares DPO-Ori (original bad cases only) with DPO-DDA (original bad cases plus synthetic bad cases rebalanced to the SFT distribution). This comparison varies both the total number of preference pairs and their domain distribution, so the causal attribution to distribution balancing is under-supported; however, this is a controlled-experiment confound, not a circularity, because DDA's definition does not include its measured effect and the test-set metrics are external to the training procedure. The zero-shot claim may be overstated because prompts still list held-out domain tools and slots, but that is a validity concern rather than a reduction of the result to its assumptions. The only self-citation with overlapping authorship (Feng et al., 2025) appears in a related-work sentence on DST and is not load-bearing for any derivation. No step in the paper's derivation chain reduces, by construction or by self-citation, to its own inputs.

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

The ledger shows that the paper's empirical claims rest on standard benchmark assumptions plus several paper-specific operationalizations (degradation definition, synthetic negative validity, prompt-based zero-shot). DPO hyperparameters and DDA data quantities are unreported tuning choices.

free parameters (4)
  • BLEU threshold for Response Agent preference pairs
    Appendix B: "we select good cases based on a certain threshold of BLEU" but the threshold value is never given; this hand-set cutoff determines which responses become positive examples in DPO training.
  • Number of DDA-generated bad cases per domain
    The DDA method balances rejected-data distribution by generating synthetic negatives, but the paper never states how many are created; this quantity is a tuning choice and is confounded with the DPO training-set size.
  • DPO hyperparameters (beta, learning rate, batch size, steps)
    Appendix B reports only 2 epochs on 8 A100 GPUs; without beta, LR, batch size, or step counts, the DPO results cannot be reproduced and the comparison may depend on unpublished tuning.
  • Degradation threshold = domain score below average accuracy
    Section 6.2 defines degradation as a domain score dropping below the average; this ad hoc threshold determines which rows in Table 2 are called degraded.
assumptions (5)
  • domain assumption MultiWOZ 2.2 is a valid and representative benchmark for task-oriented dialogue.
    Section 5.1 uses MultiWOZ 2.2 and its Inform, Success, BLEU, and Combine metrics as the sole basis for the headline claims.
  • domain assumption The DPO degradation mechanism described by Feng et al. (2024) and Pal et al. (2024) applies to fine-tuned lightweight LLMs in TOD.
    Section 4 motivates DDA from the cited conclusion that DPO lowers generation probability of dispreferred categories; the paper does not independently verify this mechanism in its own models.
  • ad hoc to paper Synthetically corrupted DDA negatives (random intent replacement, slot-value swaps, response-rule changes) are valid preference pairs.
    Appendix B describes these generation rules; the paper gives no analysis showing the synthetic negatives resemble real model errors or are suitable DPO negatives.
  • ad hoc to paper Leave-one-domain-out performance with domain-specific prompts measures zero-shot generalization.
    Section 6.3 removes training data for one domain but retains the full domain prompt with slot schemas and APIs, so the task is not truly unseen.
  • ad hoc to paper A domain's score dropping below the average accuracy indicates DPO degradation.
    Section 6.2 introduces this definition without justification; it is used to identify degraded domains in Table 2.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Empowering LLMs in Task-Oriented Dialogues: A Domain-Independent Multi-Agent Framework and Fine-Tuning Strategy." pith.science (2026). https://pith.science/paper/2RMAJEQH

@misc{pith2026250514299,
  author       = {Pith},
  title        = {Pith review of: Empowering LLMs in Task-Oriented Dialogues: A Domain-Independent Multi-Agent Framework and Fine-Tuning Strategy},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2RMAJEQH}},
  note         = {Machine review of arXiv:2505.14299}
}
read the original abstract

Task-oriented dialogue systems based on Large Language Models (LLMs) have gained increasing attention across various industries and achieved significant results. Current approaches condense complex procedural workflows into a single agent to achieve satisfactory performance on large-scale LLMs. However, these approaches face challenges to achieve comparable performance on fine-tuned lightweight LLMs, due to their limited capabilities in handling multiple complex logic. In this work, we design a Domain-Independent Multi-Agent Framework (DIMF), which contains Intent Classification Agent, Slot Filling Agent and Response Agent. This approach simplifies the learning complexity and enhances the generalization ability by separating the tasks into domain-independent components. In this framework, we enhance the capabilities in contextual understanding using the Direct Preference Optimisation (DPO) method, and propose a simple and effective Data Distribution Adaptation (DDA) method to mitigate degradation issues during DPO training. Experiments conducted on the MultiWOZ datasets show that our proposed method achieves a better average performance among all the baselines. Extensive analysis also demonstrates that our proposed framework exhibits excellent generalizability and zero-shot capability.

Figures

Figures reproduced from arXiv: 2505.14299 by the authors.

Figure 1
Figure 1. Different architectures of our proposed sys [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The main framework of our proposed method. The left part is the framework of our proposed DIMF. We [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The rewards of the chosen data and rejected [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: The Results of the DIMF after removing training data from a specific domain. The first sub-figure shows [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: An example of one round of the conversation between user and our DIMF. This case contains the history [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: The rewards of the chosen data and rejected [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: The rewards of the chosen data and rejected [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

41 extracted references · 9 canonical work pages

  1. [1]

    Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, and 1 others. 2023. Gpt-4 technical report. arXiv preprint arXiv:2303.08774

    arXiv 2023

  2. [2]

    Atheer Algherairy and Moataz Ahmed. 2024. A review of dialogue systems: current trends and future directions. Neural Computing and Applications, 36(12):6325--6351

    work page 2024

  3. [3]

    Namo Bang, Jeehyun Lee, and Myoung-Wan Koo. 2023. https://doi.org/10.18653/v1/2023.findings-acl.464 Task-optimized adapters for an end-to-end task-oriented dialogue system . In Findings of the Association for Computational Linguistics: ACL 2023, pages 7355--7369, Toronto, Canada. Association for Computational Linguistics

    work page doi:10.18653/v1/2023.findings-acl.464 2023

  4. [4]

    Willy Chung, Samuel Cahyawijaya, Bryan Wilie, Holy Lovenia, and Pascale Fung. 2023. Instructtods: Large language models for end-to-end task-oriented dialogue systems. arXiv preprint arXiv:2310.08885

    arXiv 2023

  5. [5]

    Xiang Deng, Yu Gu, Boyuan Zheng, Shijie Chen, Sam Stevens, Boshi Wang, Huan Sun, and Yu Su. 2024. Mind2web: Towards a generalist agent for the web. Advances in Neural Information Processing Systems, 36

    2024

  6. [6]

    Yilun Du, Shuang Li, Antonio Torralba, Joshua B Tenenbaum, and Igor Mordatch. 2023. Improving factuality and reasoning in language models through multiagent debate. arXiv preprint arXiv:2305.14325

    arXiv 2023

  7. [7]

    Duanyu Feng, Bowen Qin, Chen Huang, Zheng Zhang, and Wenqiang Lei. 2024. Towards analyzing and understanding the limitations of dpo: A theoretical perspective. arXiv preprint arXiv:2404.04626

    arXiv 2024

  8. [8]

    Yujie Feng, Zexin Lu, Bo Liu, Liming Zhan, and Xiao-Ming Wu. 2023. Towards llm-driven dialogue state tracking. arXiv preprint arXiv:2310.14970

    arXiv 2023

Show all 41 references
  1. [9]

    Zihao Feng, Xiaoxue Wang, Ziwei Bai, Donghang Su, Bowen Wu, Qun Yu, and Baoxun Wang. 2025. Improving generalization in intent detection: Grpo with reward-based curriculum sampling. arXiv preprint arXiv:2504.13592

    2025 arXiv

  2. [10]

    Dard: A multi-agent approach for task-oriented dialog systems

    Aman Gupta, Anirudh Ravichandran, Ziji Zhang, Swair Shah, Anurag Beniwal, and Narayanan Sadagopan. Dard: A multi-agent approach for task-oriented dialog systems. In NeurIPS 2024 Workshop on Open-World Agents

    2024

  3. [11]

    Wanwei He, Yinpei Dai, Yinhe Zheng, Yuchuan Wu, Zheng Cao, Dermot Liu, Peng Jiang, Min Yang, Fei Huang, Luo Si, and 1 others. 2022. Galaxy: A generative pre-trained model for task-oriented dialog with semi-supervised learning and explicit policy injection. In Proceedings of th...

    2022

  4. [12]

    Michael Heck, Nurul Lubis, Benjamin Ruppik, Renato Vukovic, Shutong Feng, Christian Geishauser, Hsien-Chin Lin, Carel van Niekerk, and Milica Ga s i \'c . 2023. Chatgpt for zero-shot dialogue state tracking: A solution or an opportunity? arXiv preprint arXiv:2306.01386

    2023 arXiv

  5. [13]

    Magdalena Kaiser, Patrick Ernst, and Gy \"o rgy Szarvas. 2024. Learning from relevant subgoals in successful dialogs using iterative training for task-oriented dialog systems. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 6236--6246

    2024

  6. [14]

    Nikitas Karanikolas, Eirini Manga, Nikoletta Samaridi, Eleni Tousidou, and Michael Vassilakopoulos. 2023. Large language models versus natural language understanding and generation. In Proceedings of the 27th Pan-Hellenic Conference on Progress in Computing and Informatics, pa...

    2023

  7. [15]

    Yangming Li, Kaisheng Yao, Libo Qin, Wanxiang Che, Xiaolong Li, and Ting Liu. 2020. https://doi.org/10.18653/v1/2020.acl-main.10 Slot-consistent NLG for task-oriented dialogue systems with iterative rectification network . In Proceedings of the 58th Annual Meeting of the Assoc...

    2020 doi

  8. [16]

    Tian Liang, Zhiwei He, Wenxiang Jiao, Xing Wang, Yan Wang, Rui Wang, Yujiu Yang, Shuming Shi, and Zhaopeng Tu. 2023. Encouraging divergent thinking in large language models through multi-agent debate. arXiv preprint arXiv:2305.19118

    2023 arXiv

  9. [17]

    Xiao Liu, Hao Yu, Hanchen Zhang, Yifan Xu, Xuanyu Lei, Hanyu Lai, Yu Gu, Hangliang Ding, Kaiwen Men, Kejuan Yang, and 1 others. 2023. Agentbench: Evaluating llms as agents. arXiv preprint arXiv:2308.03688

    2023 arXiv

  10. [18]

    Arka Pal, Deep Karkhanis, Samuel Dooley, Manley Roberts, Siddartha Naidu, and Colin White. 2024. Smaug: Fixing failure modes of preference optimisation with dpo-positive. arXiv preprint arXiv:2402.13228

    2024 arXiv

  11. [19]

    Baolin Peng, Michel Galley, Pengcheng He, Chris Brockett, Lars Liden, Elnaz Nouri, Zhou Yu, Bill Dolan, and Jianfeng Gao. 2022. Godel: Large-scale pre-training for goal-directed dialog. arXiv preprint arXiv:2206.11309

    2022 arXiv

  12. [20]

    Libo Qin, Wenbo Pan, Qiguang Chen, Lizi Liao, Zhou Yu, Yue Zhang, Wanxiang Che, and Min Li. 2023 a . End-to-end task-oriented dialogue: A survey of tasks, methods, and future directions. arXiv preprint arXiv:2311.09008

    2023 arXiv

  13. [21]

    Yujia Qin, Shihao Liang, Yining Ye, Kunlun Zhu, Lan Yan, Yaxi Lu, Yankai Lin, Xin Cong, Xiangru Tang, Bill Qian, and 1 others. 2023 b . Toolllm: Facilitating large language models to master 16000+ real-world apis. arXiv preprint arXiv:2307.16789

    2023 arXiv

  14. [22]

    Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D Manning, Stefano Ermon, and Chelsea Finn. 2023. Direct preference optimization: Your language model is secretly a reward model. Advances in Neural Information Processing Systems, 36:53728--53741

    2023

  15. [23]

    Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D Manning, Stefano Ermon, and Chelsea Finn. 2024. Direct preference optimization: Your language model is secretly a reward model. Advances in Neural Information Processing Systems, 36

    2024

  16. [24]

    John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. 2017. Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347

    2017 arXiv

  17. [25]

    Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, Y Wu, and 1 others. 2024. Deepseekmath: Pushing the limits of mathematical reasoning in open language models. arXiv preprint arXiv:2402.03300

    2024 arXiv

  18. [26]

    Weizhou Shen, Chenliang Li, Hongzhan Chen, Ming Yan, Xiaojun Quan, Hehong Chen, Ji Zhang, and Fei Huang. 2024. https://arxiv.org/abs/2401.07324 Small llms are weak tool learners: A multi-llm agent . Preprint, arXiv:2401.07324

    2024 arXiv

  19. [27]

    Haipeng Sun, Junwei Bao, Youzheng Wu, and Xiaodong He. 2023. Mars: Modeling context & state representations with contrastive learning for end-to-end task-oriented dialog. In Findings of the Association for Computational Linguistics: ACL 2023, pages 11139--11160

    2023

  20. [28]

    Xiangru Tang, Anni Zou, Zhuosheng Zhang, Ziming Li, Yilun Zhao, Xingyao Zhang, Arman Cohan, and Mark Gerstein. 2023. Medagents: Large language models as collaborators for zero-shot medical reasoning. arXiv preprint arXiv:2311.10537

    2023 arXiv

  21. [29]

    Lewis Tunstall, Edward Beeching, Nathan Lambert, Nazneen Rajani, Kashif Rasul, Younes Belkada, Shengyi Huang, Leandro von Werra, Cl \'e mentine Fourrier, Nathan Habib, and 1 others. 2023. Zephyr: Direct distillation of lm alignment. arXiv preprint arXiv:2310.16944

    2023 arXiv

  22. [30]

    Peiyi Wang, Lei Li, Liang Chen, Feifan Song, Binghuai Lin, Yunbo Cao, Tianyu Liu, and Zhifang Sui. 2023. Making large language models better reasoners with alignment. arXiv preprint arXiv:2309.02144

    2023 arXiv

  23. [31]

    Qingyang Wu, James Gung, Raphael Shu, and Yi Zhang. 2023. Diacttod: Learning generalizable latent dialogue acts for controllable task-oriented dialogue systems. arXiv preprint arXiv:2308.00878

    2023 arXiv

  24. [32]

    Heng-Da Xu, Xian-Ling Mao, Puhai Yang, Fanshu Sun, and He-Yan Huang. 2024. Rethinking task-oriented dialogue systems: From complex modularity to zero-shot autonomous agent. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Lo...

    2024

  25. [33]

    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, Junyang Lin, Kai Dang, and 22 others. 2024. Qwen2.5 technical repo...

    2024 arXiv

  26. [34]

    Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao. 2023. React: Synergizing reasoning and acting in language models. In International Conference on Learning Representations (ICLR)

    2023

  27. [35]

    Xiao Yu, Qingyang Wu, Kun Qian, and Zhou Yu. 2023. Krls: Improving end-to-end response generation in task oriented dialog with reinforced keywords learning. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 12338--12358

    2023

  28. [36]

    Xiaoxue Zang, Abhinav Rastogi, Srinivas Sunkara, Raghav Gupta, Jianguo Zhang, and Jindong Chen. 2020. Multiwoz 2.2: A dialogue dataset with additional annotation corrections and state tracking baselines. arXiv preprint arXiv:2007.12720

    2020 arXiv

  29. [37]

    Xiaoying Zhang, Baolin Peng, Kun Li, Jingyan Zhou, and Helen Meng. 2023. https://doi.org/10.18653/v1/2023.findings-emnlp.891 SGP - TOD : Building task bots effortlessly via schema-guided LLM prompting . In Findings of the Association for Computational Linguistics: EMNLP 2023, ...

    2023 doi

  30. [38]

    Zheng Zhang, Ryuichi Takanobu, Qi Zhu, MinLie Huang, and XiaoYan Zhu. 2020. Recent advances and challenges in task-oriented dialog systems. Science China Technological Sciences, 63(10):2011--2027

    2020

  31. [39]

    Xuhui Zhou, Hao Zhu, Leena Mathur, Ruohong Zhang, Haofei Yu, Zhengyang Qi, Louis-Philippe Morency, Yonatan Bisk, Daniel Fried, Graham Neubig, and 1 others. 2023. Sotopia: Interactive evaluation for social intelligence in language agents. arXiv preprint arXiv:2310.11667

    2023 arXiv

  32. [40]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list...

  33. [41]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.