Improving Tool Retrieval by Leveraging Large Language Models for Query Generation

Mohammad Kachuee; Puyang Xu; Sarthak Ahuja; Vaibhav Kumar; Xiaohu Liu

REVIEW 4 major objections 5 minor 1 cited by

Improving Tool Retrieval by Leveraging Large Language Models for Query Generation

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read LLM-written queries lift tool retrieval up to 23 points on known tools.

desk verdict The alignment loop is a real idea, but the 3-point out-of-domain win looks like it was chosen on the test set; treat that number as provisional. read the letter →

arxiv 2412.03573 v1 pith:AH4BNZ6W submitted 2024-11-17 cs.IR cs.AIcs.CL

Mohammad Kachuee , Sarthak Ahuja , Vaibhav Kumar , Puyang Xu , Xiaohu Liu This is my paper · ORCID

classification cs.IRcs.AIcs.CL

keywords toolretrievalquerygenerationalignmentlearningdensein-contextrejectionsampling

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 retrieval in LLM applications is usually done by vector-searching a user request against a fixed index of API descriptions. This paper proposes to insert one extra step: have an LLM read the request and write a few short tool descriptions that serve as retrieval queries. Three ways to produce those descriptions are compared, and the best in-domain results come from supervised fine-tuning on API descriptions, while an iterative alignment method that optimizes a retrieval reward gives the best out-of-domain results. The implication is that much of the retrieval quality comes from query formulation, not from the retriever architecture.

What carries the argument

The load-bearing component is the query generator: a LLaMA-13B model prompted to output up to five tool descriptions, each under twenty words, from a user request. These descriptions are embedded with the frozen all-mpnet-base-v2 model, and the resulting ranked lists are interleaved with the ranked list of the original utterance. The most distinctive mechanism is alignment learning: for each training request, the model stochastically generates 24 drafts, each draft is scored by an MMRR-style retrieval reward computed against the fixed index, the top-scoring drafts are kept after filtering, and the model is fine-tuned on those drafts; the loop repeats for five iterations.

What would settle it

Re-run the out-of-domain comparison with a different frozen embedding model (e.g., a larger sentence transformer or an instruction-tuned encoder). If the alignment method's gain over the utterance baseline shrinks or disappears under a different embedding space, the central claim that query generation is the driver of the improvement is undermined. A second check: replace the dense index with BM25 over the same descriptions; if the generated queries still outperform raw utterances, the benefit is genuinely in the query text.

Watch

Extended reading notes

Core claim

The paper's central claim is that decomposing a user request into a set of LLM-generated tool descriptions, then embedding those descriptions with a fixed sentence encoder and running cosine nearest-neighbor search over API descriptions, yields large retrieval gains over using the raw utterance. On the in-domain test set, supervised fine-tuning improves Recall@5 from 63.82% to 87.29%; on the out-of-domain test set, alignment learning improves Recall@5 from 75.26% to 78.53% while also improving MMRR and MAP. The authors interpret the alignment result as evidence that teaching the LLM to generate queries that maximize retrieval reward transfers better to unseen tools than forcing it to reproduce developer-written API descriptions.

Load-bearing premise

The approach assumes the fixed all-mpnet-base-v2 embedding space places LLM-generated tool descriptions in the same semantic neighborhood as developer-written API descriptions, so that cosine nearest-neighbor search over a frozen index is a faithful retrieval function.

Editorial extensions

If this is right

Retrieval pipelines can keep a frozen embedding index and adapt to new tool domains by retraining only the query-generation LLM.
Alignment learning provides a way to improve out-of-domain tool retrieval without new API data, as long as a training set of request-API pairs exists.
Appending the original utterance to the generated query set is a simple additive boost, especially for zero-shot generation and out-of-domain cases.
The temperature of generation matters: zero-shot works best at 1.3, SFT at 0.6, and alignment at 0.1, suggesting the alignment-trained model becomes more deterministic as it specializes.

Reading between the lines

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

If the same gain appears with a different frozen encoder, the method would be a portable pre-processing step; if not, the results are partly an artifact of all-mpnet-base-v2's embedding geometry, which the paper does not test.
The hallucinated API names in generated queries (e.g., PartyPlanningAPI) could be exploited as a data-augmentation signal or as a failure mode to suppress; even though the paper reports no notable regression, the qualitative examples show the model is composing names rather than retrieving from the catalog.
The alignment objective could be re-run with retrieval reward replaced by end-to-end task success (tool call correctness), which would likely change which drafts are kept and may yield different cross-domain behavior.

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.

Desk Editor's Note

read the letter

Short version: the paper has a genuinely useful idea—use the LLM to rewrite user requests into tool queries, then optimize that rewriting against retrieval reward via rejection sampling—but the headline out-of-domain result, a 3.27-point Recall@5 gain, is not yet credible because the authors picked temperature and alignment iterations by looking at the test set. The in-domain SFT result is much larger and not in doubt.

What's new: alignment learning for query generation, where you generate multiple drafts, score them by retrieval reward on training requests, and fine-tune on the best drafts, iterating. That is a clean way to train an LLM to produce retrieval-friendly queries for unseen tools without retraining the retriever. The paper also does real ablation work: adding the original utterance back, changing the reward metric, filter settings, and intent vs. description generation. The qualitative examples show the model decomposing requests sensibly. The SFT gains on in-domain tools are large (63.82% → 87.29% Recall@5), which is consistent and meaningful.

Soft spots. The abstract says LLM query generation improves both in-domain and out-of-domain retrieval, but zero-shot prompting is worse than the utterance baseline on both test sets (e.g., 71.76% vs 75.26% OOD Recall@5). So the claim only holds for trained methods. No variance or significance is reported anywhere. Temperature calibration (Appendix A.1) and the choice of T=5 alignment iterations (Section 4.4) both appear to be based on Recall@5 on the evaluation sets. That means the OOD alignment number is the best of a small search over temperatures and iterations, not an unbiased estimate. You need a held-out validation split before trusting the 3.27-point margin. The fixed embedding space is a secondary concern: no experiment checks whether generated queries are in-distribution for all-mpnet-base-v2, so part of the gain could be an artifact of the embedding model. And there's no comparison to HyDE or other LLM rewrites, which are the natural baselines for this exact idea.

Bottom line: the alignment method deserves attention, and the paper is worth refereeing. But the evaluator should insist on a real validation split, confidence intervals, and those baselines. As is, I'd cite the method but not the OOD number.

Referee Report

4 major / 5 minor

Summary. The paper proposes to improve dense tool retrieval by using an LLM to generate retrieval queries from user utterances, embedding those generated queries, and then performing nearest-neighbor search over an index of API descriptions. Three query-generation approaches are compared: zero-shot prompting, supervised fine-tuning on tool descriptions, and an iterative alignment-learning method that optimizes a retrieval reward via rejection sampling. Experiments on a ToolLLM-derived dataset report in-domain and out-of-domain retrieval metrics. The main empirical claims are that supervised fine-tuning gives large in-domain gains and that alignment learning gives the best out-of-domain Recall@5, while zero-shot prompting is generally worse than using the utterance directly.

Significance. If the out-of-domain alignment result holds, the paper offers a practical and conceptually simple recipe for improving tool retrieval without retraining the retriever: keep the embedding index fixed and train the query generator against a retrieval reward. The paper includes useful ablations (adding the utterance to the query set, rejection-sampling filter settings, reward-metric choice, and intent-versus-description generation) and qualitative examples that illustrate the intended behavior. The in-domain SFT gains are large and consistent. However, the headline out-of-domain advantage currently rests on a small margin (3.27 Recall@5 points) that appears vulnerable to test-set-based model selection and is reported without variance estimates, so the significance is conditional on those issues being resolved.

major comments (4)

[Abstract; Table 1] The abstract's blanket claim that LLM-generated queries improve retrieval for in-domain and out-of-domain settings is not supported by Table 1: the zero-shot method is below the utterance baseline on Recall@5 in both settings (57.86% vs. 63.82% in-domain; 71.76% vs. 75.26% out-of-domain), and SFT's out-of-domain Recall@5 advantage (76.18% vs. 75.26%) is accompanied by lower MMRR and MAP (0.6130 vs. 0.6290 and 0.6893 vs. 0.7031, respectively). The claim should be restricted to the alignment method, or the stated results should be revised.
[Section 4.3, Section 4.4, Appendices A.1 and A.3] The out-of-domain advantage of alignment appears to be selected on the test set. Appendix A.1 reports temperature calibration by measuring Recall@5 without specifying a held-out split; Appendix A.3 reports that T=5 was chosen as the best alignment setting; and Section 4.4 states that the best out-of-domain performance is reached after 5 iterations, matching Figure 3's out-of-domain peak. If these choices were made using the same 4,451 out-of-domain test examples, the reported 78.53% vs. 75.26% Recall@5 gain is the maximum of a small search over temperatures and iterations, not an unbiased estimate. Please add a proper validation split or otherwise demonstrate that model selection did not use the test labels, and report variance (e.g., confidence intervals or multiple seeds) for the headline comparison.
[Equation (2)] Equation (2) is internally inconsistent with its prose explanation. The text says the numerator is the average rank position for perfect retrieval of n items, which is (n+1)/2, but the formula as typeset uses n/2. For n=1, a perfect retrieval at rank 1 would yield 0.5 rather than 1. Because MMRR is used as the alignment reward (Appendix B.3), the definition needs to be corrected or clarified; otherwise the objective being optimized is unclear.
[Section 4.2 / Retriever Setup] The evaluation fixes the embedding model to all-mpnet-base-v2 and never varies it. The central mechanism assumes that LLM-generated query descriptions and developer-written API descriptions are close in this particular embedding space; without at least one alternative embedding model (or an analysis of query-embedding distribution shift), it is unclear whether the measured gains reflect better query wording or an artifact of the fixed encoder. Please add a sensitivity experiment or discuss this limitation explicitly.

minor comments (5)

[Appendix A.3] The phrase 'rdraf t' appears to be a typo for the variable n_draft defined in Algorithm 1; please harmonize the notation.
[Section 3.1] The sentence 'we found it it be less inclined to hallucination' contains a typo; it should read 'we found it to be less inclined to hallucinate'.
[Figure 4] The caption reads 'the distribution of at each iteration'; the word 'rewards' is missing.
[Table 1] The column header 'No Gen. LLM-Gen.' is awkward; please restructure as separate header levels, e.g., 'Baseline' and 'LLM-Gen'.
[Appendix C] The sentence 'Delta value are reported compared to the tool description generation approach' should be 'Delta values are reported'.

Circularity Check

0 steps flagged · score 0.0 of 10

No construction-level circularity: query generation never sees test labels, retrieval rewards are computed on the training set, and the central comparison is externally evaluated on held-out tools.

full rationale

The claimed derivation (LLM-generated queries improve retrieval) is not equivalent to its inputs. The retriever is a fixed all-mpnet-base-v2 encoder with cosine nearest-neighbor search (Section 4.2), and the query generator receives only user requests or, for SFT/alignment learning, training request–relevant-API pairs. The alignment reward is computed by simulating retrieval on the training set: Section 3.3 says 'we simulate retrieval of items in the train set using the generated queries in bZ1..m and compute retrieval reward for all samples,' and Algorithm 1 uses only the training inputs (X, Y). Section 4.1 explicitly states: 'The test datasets are only used for evaluation.' The in-domain/out-of-domain split is performed by tool names with overlapping samples removed, so the out-of-domain numbers are not determined by the training-set fits. The only genuinely self-referential element is hyperparameter selection: Appendix A.1 calibrates temperature by measuring Recall@5, Appendix A.3 selects T=5, and Section 4.4 says 'the best out-of-domain performance is reached after 5 iterations'; if these choices were made on the OOD test set, the reported margin may be an optimistic selection rather than an unbiased prediction, but that is a soundness threat, not a circular reduction by construction. Appendix A.3 also acknowledges the search was not a complete grid due to computational cost. The one self-citation (Kachuee et al., 2022) merely notes traditional intent classification and is not load-bearing. No equation, fitted parameter, or cited theorem reduces the central claim to its own input.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The method's success depends on several empirical assumptions and tuned hyperparameters rather than on derived equations or new physical entities. There are no invented entities and no load-bearing self-citations; the main external support is the public ToolLLM benchmark and a frozen embedding model.

free parameters (3)

Sampling temperature per generation method = zero-shot 1.3, SFT 0.6, alignment 0.1
Chosen by Recall@5 calibration in Appendix A.1; affects output diversity and can change ranking outcomes.
Alignment filter and loop settings = MMRR reward, ptop=100, rmin=0.05, ndraft=1, m=24, T=5
Selected through a limited one-at-a-time search in Appendices A.3 and B.2; these choices control which generated drafts become training targets.
Dataset preprocessing thresholds = description length 5-50 words, at most 3 relevant APIs, no overlapping labeled samples between splits
Manual filtering of ToolLLM in Section 4.1; the absolute metric values depend on this filtered split.

assumptions (4)

domain assumption ToolLLM relevance labels are correct and complete.
All evaluation metrics in Section 2.2 assume the labeled relevant APIs are exhaustive and accurate; noisy labels would distort comparative scores.
domain assumption The frozen all-mpnet-base-v2 embedding space is a valid semantic bridge between generated queries and API descriptions.
The retriever is fixed in Section 4.2 with no adaptation or contrastive training, so the method's success depends on this compatibility.
domain assumption The hand-crafted output parser recovers the intended tool descriptions from noisy LLM output.
Section 3.1 notes formatting problems and a heuristic parser is required; if the parser drops or corrupts queries, retrieval results change.
domain assumption The pre-trained LLaMA-13B has sufficient common sense to generate tool descriptions for unseen requests.
Both zero-shot and alignment rely on the base model's ability to decompose requests into API-like descriptions without having seen the API catalog, as described in Sections 3.1 and 3.3.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Improving Tool Retrieval by Leveraging Large Language Models for Query Generation." pith.science (2026). https://pith.science/paper/AH4BNZ6W

@misc{pith2026241203573,
  author       = {Pith},
  title        = {Pith review of: Improving Tool Retrieval by Leveraging Large Language Models for Query Generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AH4BNZ6W}},
  note         = {Machine review of arXiv:2412.03573}
}

read the original abstract

Using tools by Large Language Models (LLMs) is a promising avenue to extend their reach beyond language or conversational settings. The number of tools can scale to thousands as they enable accessing sensory information, fetching updated factual knowledge, or taking actions in the real world. In such settings, in-context learning by providing a short list of relevant tools in the prompt is a viable approach. To retrieve relevant tools, various approaches have been suggested, ranging from simple frequency-based matching to dense embedding-based semantic retrieval. However, such approaches lack the contextual and common-sense understanding required to retrieve the right tools for complex user requests. Rather than increasing the complexity of the retrieval component itself, we propose leveraging LLM understanding to generate a retrieval query. Then, the generated query is embedded and used to find the most relevant tools via a nearest-neighbor search. We investigate three approaches for query generation: zero-shot prompting, supervised fine-tuning on tool descriptions, and alignment learning by iteratively optimizing a reward metric measuring retrieval performance. By conducting extensive experiments on a dataset covering complex and multi-tool scenarios, we show that leveraging LLMs for query generation improves the retrieval for in-domain (seen tools) and out-of-domain (unseen tools) settings.

Figures

Figures reproduced from arXiv: 2412.03573 by the authors.

**Figure 2.** Prompt format used for the tool description [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Comparison of Recall@5 performance for the [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Bar plots showing the distribution of rewards [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

**Figure 5.** Figure 5: Prompt format used for the intent generation [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

MassTool: A Multi-Task Search-Based Tool Retrieval Framework for Large Language Models
cs.IR 2025-07 conditional novelty 6.0 of 10

A multi-task retriever that combines tool-usage detection with query-centered graph and search-based modules improves tool retrieval accuracy over prior baselines.

Reference graph

Works this paper leans on

23 extracted references · 3 canonical work pages · cited by 1 Pith paper

[1]

Yuntao Bai, Saurav Kadavath, Sandipan Kundu, Amanda Askell, Jackson Kernion, Andy Jones, Anna Chen, Anna Goldie, Azalia Mirhoseini, Cameron McKinnon, et al. 2022. Constitutional ai: Harmlessness from ai feedback. arXiv preprint arXiv:2212.08073

arXiv 2022
[2]

S \'e bastien Bubeck, Varun Chandrasekaran, Ronen Eldan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee, Yin Tat Lee, Yuanzhi Li, Scott Lundberg, et al. 2023. Sparks of artificial general intelligence: Early experiments with gpt-4. arXiv preprint arXiv:2303.12712

arXiv 2023
[3]

Jishnu Ray Chowdhury, Yong Zhuang, and Shuyi Wang. 2022. Novelty controlled paraphrase generation with retrieval augmented conditional prompt tuning. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 36, pages 10535--10544

2022
[4]

Vojt e ch Hude c ek and Ond r ej Du s ek. 2023. Are large language models all you need for task-oriented dialogue? In Proceedings of the 24th Meeting of the Special Interest Group on Discourse and Dialogue, pages 216--228

work page 2023
[5]

Gautier Izacard, Mathilde Caron, Lucas Hosseini, Sebastian Riedel, Piotr Bojanowski, Armand Joulin, and Edouard Grave. 2021. Unsupervised dense information retrieval with contrastive learning. arXiv preprint arXiv:2112.09118

arXiv 2021
[6]

Jeff Johnson, Matthijs Douze, and Herv \'e J \'e gou. 2019. Billion-scale similarity search with gpus. IEEE Transactions on Big Data, 7(3):535--547

2019
[7]

Mohammad Kachuee, Jinseok Nam, Sarthak Ahuja, Jin-Myung Won, and Sungjin Lee. 2022. Scalable and robust self-learning for skill routing in large-scale conversational ai systems. In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies: Industry Track, pages 1--8

work page 2022
[8]

Minghao Li, Feifan Song, Bowen Yu, Haiyang Yu, Zhoujun Li, Fei Huang, and Yongbin Li. 2023 a . Api-bank: A benchmark for tool-augmented llms. arXiv preprint arXiv:2304.08244

arXiv 2023

Show all 23 references

[9]

Xiaonan Li, Kai Lv, Hang Yan, Tianyang Lin, Wei Zhu, Yuan Ni, Guotong Xie, Xiaoling Wang, and Xipeng Qiu. 2023 b . Unified demonstration retriever for in-context learning. arXiv preprint arXiv:2305.04320

2023 arXiv
[10]

Shishir G Patil, Tianjun Zhang, Xin Wang, and Joseph E Gonzalez. 2023. Gorilla: Large language model connected with massive apis. arXiv preprint arXiv:2305.15334

2023 arXiv
[11]

Yujia Qin, Shengding Hu, Yankai Lin, Weize Chen, Ning Ding, Ganqu Cui, Zheni Zeng, Yufei Huang, Chaojun Xiao, Chi Han, et al. 2023 a . Tool learning with foundation models. arXiv preprint arXiv:2304.08354

2023 arXiv
[12]

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

2023 arXiv
[13]

Dragomir R Radev, Hong Qi, Harris Wu, and Weiguo Fan. 2002. Evaluating web-based question answering systems. In LREC. Citeseer

2002
[14]

Nils Reimers and Iryna Gurevych. 2019. Sentence-bert: Sentence embeddings using siamese bert-networks. arXiv preprint arXiv:1908.10084

2019 arXiv
[15]

Ohad Rubin, Jonathan Herzig, and Jonathan Berant. 2021. Learning to retrieve prompts for in-context learning. arXiv preprint arXiv:2112.08633

2021 arXiv
[16]

Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timoth \'e e Lacroix, Baptiste Rozi \`e re, Naman Goyal, Eric Hambro, Faisal Azhar, et al. 2023. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971

2023 arXiv
[17]

Andrew Yates, Rodrigo Nogueira, and Jimmy Lin. 2021. Pretrained transformers for text ranking: Bert and beyond. In Proceedings of the 14th ACM International Conference on web search and data mining, pages 1154--1156

2021
[18]

Wayne Xin Zhao, Jing Liu, Ruiyang Ren, and Ji-Rong Wen. 2022. Dense text retrieval based on pretrained language models: A survey. arXiv preprint arXiv:2211.14876

2022 arXiv
[19]

Mu Zhu. 2004. Recall, precision and average precision. Department of Statistics and Actuarial Science, University of Waterloo, Waterloo, 2(30):6

2004
[20]

URL: " 'urlintro :=

ENTRY address author booktitle chapter edition editor howpublished institution journal key month note number organization pages publisher school series title type volume year eprint doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before...
[21]

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...
[22]

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...
[23]

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 12, 2026 · model on record in the stance chip above.

[1] [1]

Yuntao Bai, Saurav Kadavath, Sandipan Kundu, Amanda Askell, Jackson Kernion, Andy Jones, Anna Chen, Anna Goldie, Azalia Mirhoseini, Cameron McKinnon, et al. 2022. Constitutional ai: Harmlessness from ai feedback. arXiv preprint arXiv:2212.08073

arXiv 2022

[2] [2]

S \'e bastien Bubeck, Varun Chandrasekaran, Ronen Eldan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee, Yin Tat Lee, Yuanzhi Li, Scott Lundberg, et al. 2023. Sparks of artificial general intelligence: Early experiments with gpt-4. arXiv preprint arXiv:2303.12712

arXiv 2023

[3] [3]

Jishnu Ray Chowdhury, Yong Zhuang, and Shuyi Wang. 2022. Novelty controlled paraphrase generation with retrieval augmented conditional prompt tuning. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 36, pages 10535--10544

2022

[4] [4]

Vojt e ch Hude c ek and Ond r ej Du s ek. 2023. Are large language models all you need for task-oriented dialogue? In Proceedings of the 24th Meeting of the Special Interest Group on Discourse and Dialogue, pages 216--228

work page 2023

[5] [5]

Gautier Izacard, Mathilde Caron, Lucas Hosseini, Sebastian Riedel, Piotr Bojanowski, Armand Joulin, and Edouard Grave. 2021. Unsupervised dense information retrieval with contrastive learning. arXiv preprint arXiv:2112.09118

arXiv 2021

[6] [6]

Jeff Johnson, Matthijs Douze, and Herv \'e J \'e gou. 2019. Billion-scale similarity search with gpus. IEEE Transactions on Big Data, 7(3):535--547

2019

[7] [7]

Mohammad Kachuee, Jinseok Nam, Sarthak Ahuja, Jin-Myung Won, and Sungjin Lee. 2022. Scalable and robust self-learning for skill routing in large-scale conversational ai systems. In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies: Industry Track, pages 1--8

work page 2022

[8] [8]

Minghao Li, Feifan Song, Bowen Yu, Haiyang Yu, Zhoujun Li, Fei Huang, and Yongbin Li. 2023 a . Api-bank: A benchmark for tool-augmented llms. arXiv preprint arXiv:2304.08244

arXiv 2023