Pith. sign in

REVIEW 4 major objections 3 minor 33 references

CRMAgent: A Multi-Agent LLM System for E-Commerce CRM Message Template Generation

T0 review · 4 major / 3 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read CRMAgent is a four-agent LLM system that rewrites weak e-commerce CRM messages, and the paper reports it beats merchants' originals on LLM-judged audience fit and marketing effectiveness.

desk verdict Useful industrial system, but the only evidence for its headline claim is the same LLM that writes the copy judging the copy. read the letter →

arxiv 2507.08325 v2 pith:PHFEJZTK submitted 2025-07-11 cs.CL cs.MA

classification cs.CLcs.MA
keywords CRMmessagetemplatesmulti-agentLLMsysteme-commerceprivate-domainmessagingretrieval-augmentedgenerationtemplaterewritingaudiencesegmentationevaluationmarketingcopy
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

CRMAgent is a multi-agent LLM system for rewriting underperforming e-commerce CRM messages, such as push notifications and in-app email, into more persuasive templates. The paper's central claim is that, across 3,957 message pairs in 11 audience segments, the generated templates beat the merchants' original templates: the LLM judge scores audience fit 9.09% higher, marketing effectiveness 38.44% higher, and prefers the generated version in 78.44% of blinded comparisons. The system is designed so that merchants without copywriting expertise could automate the rewrite using their own successful messages, similar cross-merchant templates, or a fallback rewrite when no reference exists. A sympathetic reader would care because the method is concrete and the reported gains are sizable, though the evaluation rests entirely on LLM judgments with no human or behavioral data reported.

What carries the argument

The load-bearing mechanism is the four-agent decomposition with a three-path generation router inside TemplateAgent. In path one, group-based learning uses a merchant's own top-quartile messages from the same audience segment as exemplars; in path two, retrieval-and-adaptation fetches cross-merchant templates matched on audience segment, product category, and voucher type using dense embeddings from a distilled sentence-embedding model indexed for fast similarity search; in path three, a rule-based zero-shot rewrite handles cases with no usable references. EvaluateAgent is the outcome instrument: it produces the three-level audience-match and marketing-effectiveness scores and the blinded pairwise preference that constitute the paper's evidence. In the reported configuration, ContentAgent, TemplateAgent, and EvaluateAgent all run on GPT-4o, so the same model family both writes the copy and grades it.

What would settle it

Run a randomized A/B test on the same 11 audience segments, sending the original template to one set of customers and the generated template to another, and compare seven-day click-through and conversion rates; if the generated templates do not produce higher engagement, or if human raters prefer the originals, the claimed marketing-effectiveness gains are not demonstrated.

Watch

Extended reading notes

Core claim

The paper claims that a task-specialized multi-agent LLM pipeline can turn weak e-commerce CRM templates into stronger ones. CRMAgent decomposes the work into four roles: ContentAgent compares low- and high-performing templates within an audience segment to explain why some messages engage better; RetrievalAgent searches for high-performing cross-merchant templates with matching audience label, product category, and voucher type; TemplateAgent rewrites the original message using either in-group exemplars, retrieved exemplars, or a rule-based prompt; EvaluateAgent scores both versions on audience match and marketing effectiveness or selects the preferred version without knowing which is original. The empirical payload is the head-to-head result: generated messages score 4.56 versus 4.18 on audience fit (a 9.09% relative gain) and 4.61 versus 3.33 on marketing effectiveness (a 38.44% relative gain), and EvaluateAgent prefers the generated version in 78.44% of blind comparisons. The paper adds BERTScore-F1 of 0.87 and chrF of 24.22 to show that rewrites keep the original intent while changing the wording. The authors present this as evidence that multi-agent LLM systems can serve as scalable CRM content support.

Load-bearing premise

The load-bearing premise is that the same GPT-4o model family that rewrites the messages also judges them reliably, because no human raters, click-through rates, conversions, or live A/B results are reported anywhere in the paper.

Editorial extensions

If this is right

  • If CRMAgent works as reported, a merchant without a copywriting team can upload underperforming templates and audience labels and receive a rewritten template plus an explanation of what was wrong with the original and why the new version fits the segment better.
  • The retrieval path makes the system improve as the shared template pool grows: more high-performing cross-merchant examples means more campaigns get exemplar-guided rewrites instead of the rule-based fallback.
  • Because the largest market-score gains appear where baseline scores are lowest, such as potential new customers with a 52.23% relative gain, the biggest practical payoff would be in acquisition-oriented segments, while mature segments like frequent buyers would gain less.
  • The reported BERTScore-F1 of 0.87 and chrF of 24.22 suggest the rewrite can be paired with an automatic similarity guardrail to detect when a generated template drifts from the original offer or intent.

Reading between the lines

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

  • The decisive test the paper does not run is behavioral: a live A/B test measuring click-through, conversion, or unsubscribe rates on original versus generated templates would show whether the LLM judge's 78.44% preference transfers to real customers; until then, the practical claim should be read as 'preferred by an LLM evaluator' rather than 'preferred by customers'.
  • A straightforward extension would be a human evaluation panel drawn from the target audience segments scoring the same anonymized pairs; if human preference lands well below 78.44%, that would indicate the judge is rewarding the stylistic features its own prompt asks for, such as urgency and concrete discounts, rather than generally persuasive copy.
  • Because ContentAgent, TemplateAgent, and EvaluateAgent share the same GPT-4o family, the diagnosis and the grade are not independent; rerunning evaluation with a different model or human raters would make the reported improvements more credible.
  • The audience-score dips for price-drop seekers, repeat buyers, and new followers suggest the rewrite policy trades off targeting precision for urgency; an extension would be per-segment rewrite policies that preserve price-sensitivity cues where they are load-bearing.
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 / 3 minor

Summary. The paper introduces CRMAgent, a multi-agent LLM system that rewrites underperforming e-commerce CRM message templates. It combines a ContentAgent for diagnosing template weaknesses, a RetrievalAgent for finding similar high-performing exemplars, a TemplateAgent for rewriting via group-based learning, retrieval-and-adaptation, or rule-based fallback, and an EvaluateAgent for scoring and pairwise preference judgments. The authors report experiments on a dataset derived from 3 million CRM message logs across 11 audience segments, claiming that generated templates consistently outperform merchant originals, with a 9.09% relative increase in audience-match score, a 38.44% relative increase in marketing-effectiveness score, and a 78.44% preference rate from the evaluator.

Significance. If the central claim were supported by valid evidence, CRMAgent would be a practically useful system for e-commerce CRM content generation, and the paper's modular multi-agent design and the use of segment-specific exemplars are reasonable ideas. The paper is transparent about its prompts and architecture, provides an ablation study across different backbone models, and reports segment-level results that allow readers to see variance. However, the empirical foundation is not valid for the claims made: the only quality signal comes from an LLM judge (GPT-4o) from the same model family that generates the rewrites, and no human ratings, click-through, conversion, or live A/B metrics are provided. The reported 'significant gains' are therefore unvalidated preferences of the generator's own model, not evidence about real message quality or customer behavior. The paper's own ablation and appendix statements further undermine the robustness of the evaluation. As it stands, the paper does not deliver on its advertised contribution of demonstrating practical impact.

major comments (4)
  1. [§4.1, Figures 5–6] The central claim that CRMAgent 'consistently outperforms' merchant originals rests entirely on EvaluateAgent, which is implemented with GPT-4o—the same model family used by TemplateAgent to generate the rewrites. The scoring and comparison prompts (Figures 5 and 6) explicitly reward urgency, clarity, incentive framing, and strong calls to action, which are exactly the features TemplateAgent is prompted to add (Figures 3 and 4). Thus the 78.44% preference rate and the score gains in Table 1 measure how well the generated text matches the evaluator's stylistic preferences, not whether real customers find the messages more relevant or persuasive. No human ratings, click-through rates, conversions, or A/B results are reported anywhere. This is a circular evaluation and does not support the abstract's claim of 'significant gains in both audience-match and marketing-effectiveness metrics.'
  2. [Table 1] The claim of 'consistently outperforms' is contradicted by the segment-level audience scores: audience score declines in 5 of 11 segments (Unpaid Order Buyers −1.21%, Post-Purchase Group −3.46%, Price-Drop Group −4.01%, Active Old Followers −4.19%, Repeat Buyers −4.08%) and is unchanged for Frequent Buyers. The overall +9.09% is driven mainly by two early-stage segments (Potential New Customers +41.87% and New Buyers +20.66%). An aggregate relative change over a non-uniform distribution of segment sizes does not justify the word 'consistently,' and the paper does not provide significance tests or confidence intervals for these differences.
  3. [§4.2, Table 2] The ablation study shows that the main result is evaluator-dependent: when EvaluateAgent is switched to DeepSeek-R1, the preference rate drops from 78.44% to 55.40% while the score improvements are larger. The paper attributes this to a 'potential mismatch between its scoring criteria and human-aligned quality judgments,' but no human baseline is supplied. This admission that the GPT-4o evaluator's criteria may not be human-aligned undermines the validity of the primary evaluation, since the entire outperformance claim depends on that specific evaluator.
  4. [Appendix B.2] The comparison evaluation is not position-balanced. Appendix B.2 states that 'A and B correspond to the original and generated versions respectively,' meaning Message A is always the original and Message B is always the generated message. The prompt in Figure 6 claims the evaluator does not know which is which, but because the assignment is fixed, any positional bias in the LLM judge is fully confounded with the generation effect. The paper should counterbalance the presentation order across pairs, or at minimum report a position-controlled analysis.
minor comments (3)
  1. [§3.4, Figure 5] The scoring scale is described as three levels (1, 3, 5), which precludes intermediate ratings; this coarse scale may inflate observed score differences and should be justified or replaced with a finer scale.
  2. [§4.1, Table 1] The paper does not report any measure of inter-evaluator agreement or reliability for EvaluateAgent, and it does not report the variance of the scores; without this, the aggregate Δ (%) values are hard to interpret.
  3. [Appendix C] The error-type analysis in Appendix C is performed by o3, another LLM, and the 'custom evaluation sheet' is provided in Appendix D, but there is no validation of this automated error labeling against human labels; this is an additional unvalidated LLM-as-judge component.

Circularity Check

2 steps flagged · score 7.0 of 10

Central outperformance claim rests on self-evaluation: the same GPT-4o both writes the rewrites and scores them on the exact criteria it was told to optimize.

  1. self definitional [Section 4.1, Experimental Results (GPT-4o configuration; Table 1)]
    "We use GPT-4o [15] as the underlying model for ContentAgent, TemplateAgent, and EvaluateAgent in this experiment. Table 1 shows that the generated templates consistently outperform the originals across most audience segments."

    The paper's central quantitative claim - 'generated templates consistently outperform the originals' - is measured entirely by EvaluateAgent, which is the same GPT-4o model family that TemplateAgent used to produce the rewrites. The audience scores, market scores, and preference rates in Table 1 are thus the generator's self-assessment, not an independent measurement of message quality. The claimed 9.09% and 38.44% gains reduce to GPT-4o rating its own output against the merchant original.

  2. self definitional [Section 3.3 (Figure 3) and Section 3.4 (Figures 5-6)]
    "TemplateAgent: 'rewrite the original message into a short, action-driven, persuasive push notification... Focus on improving clarity, structure, and incentive framing.' EvaluateAgent: 'Marketing Effectiveness - Is the message clear, persuasive, and motivating?' and 'Which message is more persuasive and better suited for the given audience segment?'"

    The scoring and comparison prompts ask the evaluator to reward exactly the properties that the generation prompt instructs the model to add: clarity, persuasion, action-driven wording, and incentive framing. Because the same model family executes both prompts, the 78.44% preference rate measures whether the rewrite matches its own stylistic instructions, not whether customers or any external rater find it more effective. The evaluation is therefore self-confirming by construction rather than an independent test.

full rationale

This is not a case of parameter fitting or a self-citation chain: CRMAgent's architecture is described concretely, and the authors do not lean on their own prior work to justify the system design. The circularity is in the evaluation loop. The abstract's load-bearing assertion - that CRMAgent 'consistently outperforms merchants' original templates, delivering significant gains in both audience-match and marketing-effectiveness metrics' - is supported only by GPT-4o's own judgments of texts that GPT-4o itself wrote. The generation prompt (Figure 3) and the evaluation prompt (Figures 5-6) name the same stylistic criteria, so the evaluator is checking compliance with the generator's instructions rather than measuring externally validated quality. The paper's own ablation (Table 2) strengthens this concern: when EvaluateAgent is switched to DeepSeek-R1, the preference rate drops from 78.44% to 55.40%, and the paper attributes the drop to a 'potential mismatch between its scoring criteria and human-aligned quality judgments' without providing any human judgments to anchor that claim. No human raters, click-through rates, conversions, or A/B tests are reported; Appendix B.2 further reveals that position is not counterbalanced ('A and B correspond to the original and generated versions respectively'), so the 'blind' comparison is confounded with position. A score of 7 reflects that the central result is not independent of the measuring instrument, but it is not a fully definitional identity (score 8-10) because GPT-4o could in principle have preferred some originals; the structural bias, however, makes the headline comparison self-referential.

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

The central claim rests on two unvalidated proxies: the hand-assigned engagement score used to label templates as strong or weak, and the GPT-4o evaluation used to score generated versus original messages. Both are domain assumptions with no external benchmark. No invented entities are introduced.

free parameters (3)
  • engagement_score_weights = read=1; store/voucher click=3; product card click=4; CRM button click=2; unsubscribe=-5
    Hand-assigned in Appendix A to define template performance; no validation that this weighting tracks actual marketing value.
  • strong_weak_percentiles = top 25% vs bottom 25% per audience segment
    Chosen in Appendix A to select exemplars and low performers; affects which messages are rewritten and compared.
  • evaluation_scale_levels = 1, 3, 5
    Three-level scoring in Section 3.4; the coarse scale may compress or exaggerate differences, but this is less load-bearing than the other parameters.
assumptions (4)
  • domain assumption GPT-4o's EvaluateAgent scores and preferences are valid measures of message quality.
    Central to every reported gain; Section 4.1 and the prompts in Figures 5 and 6. No human evaluation or online A/B test is provided.
  • domain assumption The hand-assigned engagement score is a valid proxy for template performance.
    Appendix A defines weighted user behaviors; used to label top and bottom 25% and to select exemplars.
  • domain assumption Embedding similarity over product-category and voucher metadata retrieves genuinely useful exemplars.
    Section 3.2; no human or quantitative check that retrieved templates are actually better references than random ones.
  • domain assumption Task decomposition into four agents improves quality over a single LLM call.
    The ablation in Section 4.2 varies backbone models but never compares against a single-agent or single-prompt baseline.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CRMAgent: A Multi-Agent LLM System for E-Commerce CRM Message Template Generation." pith.science (2026). https://pith.science/paper/PHFEJZTK

@misc{pith2026250708325,
  author       = {Pith},
  title        = {Pith review of: CRMAgent: A Multi-Agent LLM System for E-Commerce CRM Message Template Generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PHFEJZTK}},
  note         = {Machine review of arXiv:2507.08325}
}
read the original abstract

In e-commerce private-domain channels such as instant messaging and e-mail, merchants engage customers directly as part of their Customer Relationship Management (CRM) programmes to drive retention and conversion. While a few top performers excel at crafting outbound messages, most merchants struggle to write persuasive copy because they lack both expertise and scalable tools. We introduce CRMAgent, a multi-agent system built on large language models (LLMs) that generates high-quality message templates and actionable writing guidance through three complementary modes. First, group-based learning enables the agent to learn from a merchant's own top-performing messages within the same audience segment and rewrite low-performing ones. Second, retrieval-and-adaptation fetches templates that share the same audience segment and exhibit high similarity in voucher type and product category, learns their successful patterns, and adapts them to the current campaign. Third, a rule-based fallback provides a lightweight zero-shot rewrite when no suitable references are available. Extensive experiments show that CRMAgent consistently outperforms merchants' original templates, delivering significant gains in both audience-match and marketing-effectiveness metrics.

Figures

Figures reproduced from arXiv: 2507.08325 by the authors.

Figure 1
Figure 1. Framework of CRMAgent, a multi-agent system for improving underperforming CRM message templates in e [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Prompt used by ContentAgent to generate compara￾tive analyses of CRM template performance. computed based on structured metadata, including the product cat￾egory hierarchy and voucher attributes. These vectors are indexed using Faiss [3] for efficient similarity search. At inference time, the agent encodes the target message’s metadata into a dense vector and retrieves the top-𝑘 most similar templates to serve as fe… view at source ↗
Figure 4
Figure 4. Prompt for TemplateAgent You are a CRM marketing copywriter. A past CRM message performed poorly. Below is the Original Poor Template: {Poor Performance Template} Problem Analysis: {Poor Performance Reason} Here are some high-performing templates from the same or similar campaigns: {Good Performance Template} Summary of what makes high-performing templates successful: {Good Performance Reason} Now, please rewrite th… view at source ↗
Figures from the paper (10 more)
Figure 3
Figure 3. Figure 3: Prompt used by TemplateAgent to rewrite underper￾forming CRM messages using performance diagnostics and exemplar templates. 3.4 EvaluateAgent EvaluateAgent assesses the quality of CRM message template revisions through two complementary evaluation modes. In the scoring…
Figure 6
Figure 6. Figure 6: Prompt for comparative evaluation, where [PITH_FULL_IMAGE:figures/full_fig_p004_6.png]
Figure 9
Figure 9. Figure 9: Distribution of CRM plans across product cate [PITH_FULL_IMAGE:figures/full_fig_p007_9.png]
Figure 8
Figure 8. Figure 8: Distribution of average engagement scores across [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 10
Figure 10. Figure 10: Illustrative example of TemplateAgent response for the potential new customer audience segment. explicit expression of appreciation and provides a clear benefit— twenty percent off the next order. This wording reinforces the value of continued engagement while making …
Figure 12
Figure 12. Figure 12: Illustrative example of EvaluateAgent response (audience match and marketing scores) for the potential new customer audience segment. Similarly, in the active old followers segment shown in [PITH_FULL_IMAGE:figures/full_fig_p008_12.png]
Figure 11
Figure 11. Figure 11: Illustrative example of TemplateAgent response for the active old followers audience segment. B.2 EvaluateAgent Responses To prevent agents from being biased by labels such as “original” or “generated”, each pair of CRM messages is anonymized as Message A and Message …
Figure 15
Figure 15. Figure 15: Illustrative example of EvaluateAgent preference decision for the active old customer audience segment. C Template Quality Analyses We conduct a systematic analysis of CRM message template quality using our custom evaluation sheet (Appendix D). Each template is assess…
Figure 13
Figure 13. Figure 13: Illustrative example of EvaluateAgent response (audience match and marketing scores) for the active old customer audience segment. GPT-4o Response by EvaluateAgent for Potential New Customers Preferred Message: B Reason: Message B is more persuasive and better suited …
Figure 14
Figure 14. Figure 14: Illustrative example of EvaluateAgent preference decision for the potential new customer audience segment. who are already loyal. The agent’s decision highlights the impor￾tance of aligning promotional language with customer lifecycle stage—what works for new users ma…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

33 extracted references · 16 canonical work pages

  1. [1]

    Yuemin Chen, Feifan Wu, Jingwei Wang, Hao Qian, Ziqi Liu, Zhiqiang Zhang, Jun Zhou, and Meng Wang. 2024. Knowledge-augmented Financial Market Analysis and Report Generation. InProceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: Industry Track . 1207–1217

  2. [2]

    Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guil- laume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettle- moyer, and Veselin Stoyanov. 2019. Unsupervised Cross-lingual Representa- tion Learning at Scale. CoRR abs/1911.02116 (2019). arXiv:1911.02116 http: //arxiv.org/abs/1911.02116

  3. [3]

    Matthijs Douze, Alexandr Guzhva, Chengqi Deng, Jeff Johnson, Gergely Szilvasy, Pierre-Emmanuel Mazaré, Maria Lomeli, Lucas Hosseini, and Hervé Jégou. 2024. The Faiss library. (2024). arXiv:2401.08281 [cs.LG]

  4. [4]

    Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, et al . 2025. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv preprint arXiv:2501.12948 (2025)

  5. [5]

    Yue Guo and Yi Yang. 2024. EconNLI: Evaluating Large Language Models on Economics Reasoning. arXiv preprint arXiv:2407.01212 (2024)

  6. [6]

    Tatsunori B Hashimoto, Kelvin Guu, Yonatan Oren, and Percy S Liang. 2018. A retrieve-and-edit framework for predicting structured outputs. Advances in Neural Information Processing Systems 31 (2018)

  7. [7]

    Saverio Ieva, Ivano Bilenchi, Filippo Gramegna, Agnese Pinto, Floriano Scioscia, Michele Ruta, and Giuseppe Loseto. 2025. Enhancing Last-Mile Logistics: AI- Driven Fleet Optimization, Mixed Reality, and Large Language Model Assistants for Warehouse Operations. Sensors 25, 9 (2025), 2696

  8. [8]

    Vineet Kumar and Werner Reinartz. 2018. Customer relationship management. Springer

Show all 33 references
  1. [9]

    Beibin Li, Konstantina Mellou, Bo Zhang, Jeevan Pathuri, and Ishai Menache

  2. [10]

    Yuhang Liu, Xueyu Hu, Shengyu Zhang, Jingyuan Chen, Fan Wu, and Fei Wu

  3. [11]

    Zefang Liu and Yinzhu Quan. 2024. TPP-LLM: Modeling Temporal Point Processes by Efficiently Fine-Tuning Large Language Models. arXiv preprint arXiv:2410.02062 (2024)

  4. [12]

    Zefang Liu and Yinzhu Quan. 2025. EconWebArena: Benchmarking Autonomous Agents on Economic Tasks in Realistic Web Environments. arXiv preprint arXiv:2506.08136 (2025)

  5. [13]

    Zefang Liu and Yinzhu Quan. 2025. Retrieval of temporal event sequences from textual descriptions. In Proceedings of the 4th International Workshop on Knowledge-Augmented Methods for Natural Language Processing . 37–49

  6. [14]

    Vijit Malik, Akshay Jagatap, Vinayak Puranik, and Anirban Majumder. 2024. PEARL: Preference extraction with exemplar augmentation and retrieval with LLM agents. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: Industry Track . 1536–1547

  7. [15]

    OpenAI. 2024. Hello GPT-4o. https://openai.com/index/hello-gpt-4o/

  8. [16]

    OpenAI. 2025. Introducing GPT-4.1 in the API. https://openai.com/index/gpt-4-1/

  9. [17]

    OpenAI. 2025. Introducing OpenAI o3 and o4-mini. https://openai.com/index/ introducing-o3-and-o4-mini/

  10. [18]

    Maja Popović. 2015. chrF: character n-gram F-score for automatic MT evaluation. In Proceedings of the tenth workshop on statistical machine translation . 392–395

  11. [19]

    Yinzhu Quan and Zefang Liu. 2024. Econlogicqa: A question-answering bench- mark for evaluating large language models in economic sequential reasoning. arXiv preprint arXiv:2405.07938 (2024)

  12. [20]

    Yinzhu Quan and Zefang Liu. 2024. Invagent: A large language model based multi-agent system for inventory management in supply chains. arXiv preprint arXiv:2407.11384 (2024)

  13. [21]

    Yinzhu Quan, Yujia Xu, Guanlin Chen, Frederick Benaben, and Benoit Montreuil

  14. [22]

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

  15. [23]

    Harmanpreet Singh, Nikhil Verma, Yixiao Wang, Manasa Bharadwaj, Homa Fashandi, Kevin Ferreira, and Chul Lee. 2024. Personal Large Language Model Agents: A Case Study on Tailored Travel Planning. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Pro...

  16. [24]

    Shenao Wang, Yanjie Zhao, Xinyi Hou, and Haoyu Wang. 2025. Large language model supply chain: A research agenda. ACM Transactions on Software Engineer- ing and Methodology 34, 5 (2025), 1–46

  17. [25]

    Wenhui Wang, Furu Wei, Li Dong, Hangbo Bao, Nan Yang, and Ming Zhou

  18. [26]

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems 35 (2022), 24824–24837

  19. [27]

    Dayu Yang, Antoine Simoulin, Xin Qian, Xiaoyi Liu, Yuwei Cao, Zhaopu Teng, and Grey Yang. 2025. DocAgent: A Multi-Agent System for Automated Code Documentation Generation. arXiv preprint arXiv:2504.08725 (2025)

  20. [28]

    Mingyue Yuan, Jieshan Chen, and Aaron Quigley. 2024. MAxPrototyper: A Multi-Agent Generation System for Interactive User Interface Prototyping. arXiv preprint arXiv:2405.07131 (2024)

  21. [29]

    Potential New Customers

    Tianyi Zhang, Varsha Kishore, Felix Wu, Kilian Q Weinberger, and Yoav Artzi. 2019. Bertscore: Evaluating text generation with bert. arXiv preprint arXiv:1904.09675 (2019). A Data We begin with over 3 million successfully delivered CRM message logs from April 2025, progressivel...

  22. [2020]

    Advances in neural information processing systems 33 (2020), 5776–5788

    Minilm: Deep self-attention distillation for task-agnostic compression of pre-trained transformers. Advances in neural information processing systems 33 (2020), 5776–5788

  23. [2023]

    arXiv preprint arXiv:2307.03875 (2023)

    Large language models for supply chain optimization. arXiv preprint arXiv:2307.03875 (2023)

  24. [2024]

    arXiv preprint arXiv:2411.03957 (2024)

    Fine-Grained Guidance for Retrievers: Leveraging LLMs’ Feedback in Retrieval-Augmented Generation. arXiv preprint arXiv:2411.03957 (2024)

  25. [2025]

    arXiv preprint arXiv:2503.21115 (2025)

    Leveraging Large Language Models for Risk Assessment in Hyperconnected Logistic Hub Network Deployment. arXiv preprint arXiv:2503.21115 (2025)

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.