Pith. sign in

REVIEW 3 major objections 3 minor 1 cited by

Prompt Engineering for Requirements Engineering: A Literature Review and Roadmap

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

Pith's one-line read This paper claims that prompt engineering for requirements engineering can be systematically organized by a hybrid taxonomy, and that mapping 35 studies onto it exposes specific gaps and a four-stage roadmap.

desk verdict A useful, transparent SLR of PE4RE with a valuable study corpus, but the roadmap rests on a gap claim contradicted by the paper's own included studies. read the letter →

arxiv 2507.07682 v1 pith:UG6CINO4 submitted 2025-07-10 cs.SE

classification cs.SE
keywords promptengineeringrequirementssystematicliteraturereviewtaxonomyroadmaplargelanguagemodelschain-of-thoughtretrieval-augmentedgeneration
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

This paper tries to bring order to the fragmented practice of using prompt engineering in requirements engineering, a field the authors call PE4RE. It reports the first roadmap-oriented systematic review of the area, screening 867 records down to 35 primary studies from 2018 to 2025. The authors propose a hybrid taxonomy that sorts prompting techniques such as few-shot priming, chain-of-thought reasoning, and retrieval-augmented generation by the requirements-engineering role they serve, and they map every study onto it. They then identify the gaps the mapping reveals and lay out a four-stage roadmap for turning today's ad-hoc prototypes into reproducible, practitioner-friendly workflows. A sympathetic reader would care because the review gives the field a common vocabulary and a concrete agenda for making LLM-based requirements work auditable and comparable.

What carries the argument

The load-bearing object is the hybrid PE taxonomy of eight categories: Creative Generation; Contextualization and Personalization; Reasoning and Step-wise Thinking; Knowledge Augmentation; Multimodal Understanding; Meta-cognition and Self-reflection; Information Extraction and Classification; and Code Generation. The taxonomy does the organizing work of the review: every primary study is assigned to one or more categories, and the resulting counts and co-occurrence patterns are the evidence base for the gap analysis. The second piece of machinery is the four-stage roadmap (R1–R4), which converts each observed gap into a research task, a starter experiment, and a deliverable, and adds a reporting checklist as the mechanism for making future studies comparable.

What would settle it

Re-run the search without the mandatory "prompt engineering" keyword, adding arXiv and grey-literature venues, and count how many additional requirements-engineering studies use prompting techniques under other names; if a substantial number of multimodal or elicitation-focused studies appear, the reported zero-count gaps would be refuted. Alternatively, have two independent researchers re-code the same 35 studies into the eight categories and measure inter-rater agreement; low agreement would invalidate the frequency and co-occurrence analysis that drives the roadmap.

Watch

Extended reading notes

Core claim

The central discovery, stated on the paper's own terms, is that the PE4RE literature is not a random collection of tricks but is organized around a small set of recurring prompt designs, and that this set can be captured by an eight-category taxonomy spanning technique-oriented and task-oriented dimensions. Applied to the 35 studies, the taxonomy shows that role-based contextualization and personalization appears in 26 studies, reasoning and stepwise thinking in 20, and that the two are usually combined, making "role prompt plus chain-of-thought" the field's dominant recipe. It also shows that knowledge augmentation, meta-cognition, and self-reflection are layered on top of reasoning rather than used as standalone paradigms, and that multimodal understanding appears in none of the 35 studies. From these patterns the paper derives three limitations—no non-textual prompting, little work on requirements elicitation, and almost no ablation-based evaluation—and answers them with a four-stage roadmap.

Load-bearing premise

The review's conclusions stand or fall on whether the 35 sampled studies fairly represent the field, and on whether the authors' manual assignment of each study to taxonomy categories is consistent and repeatable by other researchers.

Editorial extensions

If this is right

  • New PE4RE techniques will need to beat the dominant baseline of role prompt plus chain-of-thought, with or without retrieval augmentation.
  • Prompt studies in requirements engineering should include ablations; currently only 2 of 20 reasoning-based papers do, so reported gains are not attributable to specific prompt elements.
  • Multimodal prompting is an open niche: no reviewed study processes diagrams, mockups, or other non-textual inputs, despite RE artifacts being frequently visual.
  • Requirements elicitation is under-served by current prompting work, so conversational and interview-style prompt patterns are a clear next target.
  • A community benchmark and a standard reporting checklist (model, prompt template, context length, metrics) would make future PE4RE results comparable and replicable.

Reading between the lines

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

  • If the taxonomy becomes accepted, it can double as a design and reporting checklist for future papers, much as empirical reporting guidelines have shaped other fields.
  • The zero-multimodal result suggests a cheap, testable experiment: take a text-only study from the corpus, run it on a multimodal model with its diagrams or mockups included, and measure whether traceability or consistency improves.
  • The mandatory "prompt engineering" search term may hide a larger literature that uses "in-context learning," "instruction following," or "prompt-based" wording; if so, the field's true size and coverage could differ from the 35-study picture.
  • The roadmap's traceability protocol could be piloted immediately on existing datasets such as the sequence-diagram and goal-model corpora, without waiting for the full benchmark suite.
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

3 major / 3 minor

Summary. This manuscript presents a systematic literature review of prompt engineering for requirements engineering (PE4RE), following Kitchenham's and Petersen's secondary-study guidelines. The authors screened 867 records from six digital libraries, selected 35 primary studies, and propose a hybrid taxonomy that combines technique-oriented prompting categories (e.g., reasoning, contextualization, knowledge augmentation) with task-oriented RE roles. They analyze the selected studies by target task, LLM family, and prompting method, identify limitations (L1–L3), and derive a four-stage roadmap (R1–R4) for maturing PE4RE from ad-hoc prototypes to reproducible, industry-ready workflows.

Significance. If the synthesis is made internally consistent, this would be the first roadmap-oriented systematic review at the intersection of prompt engineering and requirements engineering, and its taxonomy and roadmap could become useful reference points for future work. The paper's strengths include a transparent and reproducible search protocol, a complete list of the 35 primary studies in the appendix, detailed illustrative examples for each taxonomy category (Listings 1–8), and an explicit threats-to-validity section. However, the current version contains internal inconsistencies in the gap analysis and in the quantitative mapping that undermine the roadmap's foundations as presented.

major comments (3)
  1. [§V-B (RQ2-1), L2 and §V-B (RQ2-2), R2] The claimed gap L2 states that 'none of the papers formally report requirements generation or elicitation' and that only P15, P27, P25, and P33 prompt LLMs to ask questions, conduct interviews, or simulate stakeholder interactions. This is directly contradicted by the paper's own primary-study list: Table IV lists P18 as 'Generating Req. Elicitation Interview Scripts', P23 as 'Evolving Req. Elicitation from App Reviews', and P16 as 'Multi-agent Sys. for Elicitation/Analysis'; these studies appear in Table VII with prompting methods and are categorized under 'Req. Analysis & Gen.' in Table V. Because roadmap recommendation R2 is explicitly justified as 'addresses limitation L2: elicitation under-explored', the false gap statement is load-bearing: if P18, P23, and P16 qualify as elicitation studies, the premise for R2 collapses. The authors need to correct this statement or provide a principled justification for excluding these studies from the elicitation gap analysis.
  2. [Table V and §V-A (RQ1-1)] Table V double-counts P21: it appears in the 'Software Analysis' row (P2, P20, P21, P25) and also in the 'SDLC Assistant' row (P6, P21). This yields 36 task assignments for 35 primary studies and inflates the reported frequencies for both categories. The quantitative summary in RQ1-1 relies on these counts, so the task distribution and the relative emphasis in the discussion are affected. The authors should assign P21 to a single category or explicitly report multi-label assignments with a corresponding total count.
  3. [§IV-B and Table VIII] The taxonomy is described as 'designed to be mutually exclusive yet collectively exhaustive', but the mapping in Table VIII assigns many studies to multiple categories (e.g., P1 to both Reasoning & Step-wise Thinking and Contextualization & Personalization; P21 to four categories). If the categories are intended to be mutually exclusive for a given prompting method, the multi-label assignments need to be reconciled with that claim; if the taxonomy instead permits multiple categories per study, the phrase 'mutually exclusive' should be qualified to make the scope of the exclusivity claim precise. This matters because the paper's categorization is a central contribution and the reported co-occurrence patterns build on multi-label assignments.
minor comments (3)
  1. [§III-B-2 and Table I] The search strings require a group containing 'prompt engineering' or related terms, but inclusion criterion I02 states that a paper qualifies even if it 'does not explicitly use the term prompt engineering'. This tension is acknowledged in the text, but the implications for sample representativeness should also be discussed in the threats-to-validity section, since studies that use prompting without the keyword may be systematically excluded.
  2. [Table VI] The table title says 'THE NUMBER OF LLMs ADOPTED', but the 'PLM' row includes BERT, ALBERT, RoBERTa, and XLNet, which are pre-trained language models rather than large language models in the usual sense. Either rename the table to 'language models' or clarify the inclusion criterion for this row.
  3. [§V-A (RQ1-1), bullet list] The bullet after Table IV says 'Requirements Analysis and Generation is the most common task (15 studies)', but the count in Table V is 15; this is consistent. However, the text's 'Requirements generation (P16, P18, P19, P23, P27, P30, P35)' ignores the double-counting issue mentioned above and should be cross-checked after the correction of Table V.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the survey's taxonomy and roadmap are syntheses of cited primary studies; self-citations are not load-bearing, and the L2/R2 inconsistency is a correctness/validity issue, not a circular derivation.

full rationale

This is a secondary study with no fitted parameters or first-principles derivation. The proposed taxonomy is explicitly an acknowledged consolidation of external surveys (Sahoo et al. [5] and Fagbohun et al. [6]), and the roadmap recommendations are proposals grounded in the included primary studies rather than quantities derived from those studies by construction. The only author self-citations appear as included primary studies (P7, P26) and as general references ([2], [17], [25]); removing them would not change the taxonomy structure or the roadmap's logic, so self-citation is not load-bearing. The paper is also transparent about its search-string limitation (Section III-B-2: treating 'Prompt* Engineering' as a mandatory group 'may have excluded studies') and its grey-literature exclusion (Section VI), which are acknowledged scope limits rather than hidden inputs. The more serious issue is internal inconsistency: L2 claims 'none of the papers formally report requirements generation or elicitation' and lists only P15, P27, P25, P33 as prompting LLMs to ask questions, conduct interviews, or simulate stakeholder interactions, while the paper's own Table IV lists P16 as 'Multi-agent Sys. for Elicitation/Analysis', P18 as 'Generating Req. Elicitation Interview Scripts', and P23 as 'Evolving Req. Elicitation from App Reviews'. This undermines the empirical support for R2, but it is a data-consistency/internal-validity threat, not a case where the conclusion is equivalent to its input by construction. Likewise, double-counting P21 in Table V (36 task entries for 35 papers) is an aggregation error, not circularity. Overall, the central claim is an organizational synthesis rather than a derivation, so circularity is minimal.

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

The review rests on methodological assumptions about search coverage and classification reliability. No free parameters are fitted, and the only introduced conceptual artifact is the taxonomy, which is not an empirical entity. The internal inconsistencies in the tables indicate that the classification-related axioms are not fully satisfied.

assumptions (3)
  • domain assumption The search protocol, including the mandatory prompt engineering term group, retrieves a representative sample of PE4RE studies.
    The authors state that treating 'Prompt* Engineering' as mandatory maximizes reproducibility but may exclude studies introducing novel prompting techniques without explicit mention (Section III-B-2). If this exclusion is significant, the mapping and gap analysis are incomplete.
  • ad hoc to paper The proposed taxonomy categories are mutually exclusive and collectively exhaustive for classifying prompting methods in RE.
    The paper claims this property in Section IV-B, but Table VII assigns multiple categories to individual studies and Table VIII adds two categories (Defeater Identification, Defect Correction) absent from Fig. 2, so the property is not demonstrated.
  • domain assumption Manual screening and classification by the authors, with consensus, is reliable.
    Selection and categorization were done by the first and second authors independently with cross-validation, but no inter-rater reliability measure is reported; Table V double-counts P21, suggesting the classification step is error-prone.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Prompt Engineering for Requirements Engineering: A Literature Review and Roadmap." pith.science (2026). https://pith.science/paper/UG6CINO4

@misc{pith2026250707682,
  author       = {Pith},
  title        = {Pith review of: Prompt Engineering for Requirements Engineering: A Literature Review and Roadmap},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UG6CINO4}},
  note         = {Machine review of arXiv:2507.07682}
}
read the original abstract

Advancements in large language models (LLMs) have led to a surge of prompt engineering (PE) techniques that can enhance various requirements engineering (RE) tasks. However, current LLMs are often characterized by significant uncertainty and a lack of controllability. This absence of clear guidance on how to effectively prompt LLMs acts as a barrier to their trustworthy implementation in the RE field. We present the first roadmap-oriented systematic literature review of Prompt Engineering for RE (PE4RE). Following Kitchenham's and Petersen's secondary-study protocol, we searched six digital libraries, screened 867 records, and analyzed 35 primary studies. To bring order to a fragmented landscape, we propose a hybrid taxonomy that links technique-oriented patterns (e.g., few-shot, Chain-of-Thought) to task-oriented RE roles (elicitation, validation, traceability). Two research questions, with five sub-questions, map the tasks addressed, LLM families used, and prompt types adopted, and expose current limitations and research gaps. Finally, we outline a step-by-step roadmap showing how today's ad-hoc PE prototypes can evolve into reproducible, practitioner-friendly workflows.

Figures

Figures reproduced from arXiv: 2507.07682 by the authors.

Figure 1
Figure 1. Procedure of the Study ing a software system’s scope and success [13]. Acting as a conduit between stakeholder needs and technical implementa￾tion, RE negotiates diverse perspectives to produce verifiable requirements [14], [15]. Its iterative cycle, comprising elicita￾tion, analysis, specification, validation, and ongoing manage￾ment, ensures that new insights and contexts continually refine the requirements [16], … view at source ↗
Figure 2
Figure 2. PE Categorization: Blue are technique-oriented categories, Grey enable both techniques and task-oriented, and Red are task-oriented categories. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Discussion (0). Sign in 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

  1. Multi-Modal Requirements Data-based Acceptance Criteria Generation using LLMs

    cs.SE 2025-08 conditional novelty 6.0 of 10

    Multi-modal RAG (text plus UI screenshots) with reward-based polishing generates acceptance criteria from user stories that three industry experts rated near 4/5 on relevance, correctness, and understandability.

Reference graph

Works this paper leans on

45 extracted references · 20 canonical work pages · cited by 1 Pith paper

  1. [1]

    Requirements engineering as a success factor in software projects,

    H. F. Hofmann and F. Lehner, “Requirements engineering as a success factor in software projects,”IEEE software, vol. 18, no. 4, 2001

  2. [2]

    Advancing requirements engineering through generative ai: Assessing the role of llms,

    C. Arora, J. Grundy, and M. Abdelrazek, “Advancing requirements engineering through generative ai: Assessing the role of llms,”arXiv preprint arXiv:2310.13976, 2023

  3. [3]

    Enhancing and Accelerating Large Language Models via Instruction-Aware Contextual Compression

    H. Hou, F. Ma, B. Bai, X. Zhu, and F. Yu, “Enhancing and accelerating large language models via instruction-aware contextual compression,” arXiv preprint arXiv:2408.15491, 2024

  4. [4]

    Automatically Generating Rules of Malicious Software Packages via Large Language Model

    X. Zhang, H. Chen, Y . He, W. Niu, and Q. Li, “Automatically generating rules of malicious software packages via large language model,”arXiv preprint arXiv:2504.17198, 2025

  5. [5]

    A systematic survey of prompt engineering in large language models: Techniques and applications,

    P. Sahoo, A. K. Singh, S. Saha, V . Jain, S. Mondal, and A. Chadha, “A systematic survey of prompt engineering in large language models: Techniques and applications,”arXiv preprint arXiv:2402.07927, 2024

  6. [6]

    An empirical categorization of prompting techniques for large language models: A practitioner’s guide,

    O. Fagbohun, R. M. Harrison, and A. Dereventsov, “An empirical categorization of prompting techniques for large language models: A practitioner’s guide,”arXiv preprint arXiv:2402.14837, 2024

  7. [7]

    Prompt engineering in large language models,

    G. Marvin, N. Hellen, D. Jjingo, and J. Nakatumba-Nabende, “Prompt engineering in large language models,” inInternational conference on data intelligence and cognitive informatics. Springer, 2023

  8. [8]

    Language models are few-shot learners,

    T. B. Brown, B. Mann, N. Ryder, M. Subbiah, J. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askellet al., “Language models are few-shot learners,”arXiv preprint arXiv:2005.14165, 2020

Show all 45 references
  1. [9]

    Chain-of-thought prompting elicits reasoning in large language models,

    J. Wei, X. Wang, D. Schuurmans, M. Bosma, B. Ichter, F. Xia, H. Chi, Q. V . Le, and D. Zhou, “Chain-of-thought prompting elicits reasoning in large language models,”arXiv preprint arXiv:2201.11903, 2022

  2. [10]

    Retrieval- augmented generation for knowledge-intensive nlp tasks,

    P. Lewis, E. Perez, A. Piktus, F. Petroni, V . Karpukhin, N. Goyal, H. K ¨uttler, M. Lewis, W.-t. Yih, T. Rockt ¨aschelet al., “Retrieval- augmented generation for knowledge-intensive nlp tasks,”Advances in neural information processing systems, vol. 33, 2020

  3. [11]

    Segress: Software engi- neering guidelines for reporting secondary studies,

    B. Kitchenham, L. Madeyski, and D. Budgen, “Segress: Software engi- neering guidelines for reporting secondary studies,”IEEE Transactions on Software Engineering, vol. 49, no. 3, 2022

  4. [12]

    Guidelines for conduct- ing systematic mapping studies in software engineering: An update,

    K. Petersen, S. Vakkalanka, and L. Kuzniarz, “Guidelines for conduct- ing systematic mapping studies in software engineering: An update,” Information and software technology, vol. 64, 2015

  5. [13]

    Information requirements for enterprise systems,

    I. Sommerville, R. Lock, and T. Storer, “Information requirements for enterprise systems,”ArXiv, vol. abs/1209.5246, 2012

  6. [14]

    A meta-level design science process for integrating stakeholder needs - demonstrated for smart city services,

    A. Knutas, Z. Pourzolfaghar, and M. Helfert, “A meta-level design science process for integrating stakeholder needs - demonstrated for smart city services,” inInternational Conference on Computer-Human Interaction Research and Applications, 2017

  7. [15]

    Effective requirement engineering process model in software engineering,

    I. C. Udousoro, “Effective requirement engineering process model in software engineering,”Software Engineering, 2020

  8. [16]

    Exploratory study to assess and evaluate requirement specification techniques using analysis determination requirements framework,

    S. Besrour, L. B. A. Rahim, and P. D. D. Dominic, “Exploratory study to assess and evaluate requirement specification techniques using analysis determination requirements framework,”Research Journal of Applied Sciences, Engineering and Technology, vol. 9, 2015

  9. [17]

    Requirements-driven automated software testing: A systematic review,

    F. Wang, C. Arora, C. Tantithamthavorn, K. Huang, and A. Aleti, “Requirements-driven automated software testing: A systematic review,”

  10. [18]

    Pre- train, prompt, and predict: A systematic survey of prompting methods in natural language processing,

    P. Liu, W. Yuan, J. Fu, Z. Jiang, H. Hayashi, and G. Neubig, “Pre- train, prompt, and predict: A systematic survey of prompting methods in natural language processing,”arXiv preprint arXiv:2107.13586, 2021

  11. [19]

    Generative reliability-based design optimization using in-context learning capabilities of large language models,

    Z. Jiang, Q. Tang, and Z. Wang, “Generative reliability-based design optimization using in-context learning capabilities of large language models,”ArXiv, vol. abs/2503.22401, 2025

  12. [20]

    Large language models: A survey,

    S. Minaee, T. Mikolov, N. Nikzad, M. A. Chenaghlu, R. Socher, X. Amatriain, and J. Gao, “Large language models: A survey,”ArXiv, vol. abs/2402.06196, 2024

  13. [21]

    Prompt engineering: a methodology for optimizing interactions with ai-language models in the field of engineering,

    J. D. Vel ´asquez-Henao, C. J. Franco-Cardona, and L. Cadavid-Higuita, “Prompt engineering: a methodology for optimizing interactions with ai-language models in the field of engineering,”DYNA, 2023

  14. [22]

    The prompt report: A systematic survey of prompting techniques,

    S. Schulhoff, M. Ilie, N. Balepur, K. Kahadze, A. Liu, C. Si, Y . Li, A. Gupta, H. Han, S. Schulhoffet al., “The prompt report: A systematic survey of prompting techniques,”arXiv:2406.06608, vol. 5, 2024

  15. [23]

    Teler: A general taxonomy of llm prompts for benchmarking complex tasks,

    S. K. K. Santu and D. Feng, “Teler: A general taxonomy of llm prompts for benchmarking complex tasks,”arXiv:2305.11430, 2023

  16. [24]

    Systematic literature review of prompt engineering patterns in software engineering,

    Y . Sasaki, H. Washizaki, J. Li, D. Sander, N. Yoshioka, and Y . Fukazawa, “Systematic literature review of prompt engineering patterns in software engineering,” in2024 IEEE 48th Annual Computers, Software, and Applications Conference (COMPSAC). IEEE, 2024

  17. [25]

    Generating test scenarios from nl requirements using retrieval-augmented llms: An industrial study,

    C. Arora, T. Herda, and V . Homm, “Generating test scenarios from nl requirements using retrieval-augmented llms: An industrial study,” in International Requirements Engineering Conference (RE’24), 2024

  18. [26]

    Learning to retrieve prompts for in-context learning,

    O. Rubin, J. Herzig, and J. Berant, “Learning to retrieve prompts for in-context learning,”arXiv preprint arXiv:2112.08633, 2021

  19. [27]

    Large language models understand and can be enhanced by emotional stimuli,

    C. Li, J. Wang, Y . Zhang, K. Zhu, W. Hou, J. Lian, F. Luo, Q. Yang, and X. Xie, “Large language models understand and can be enhanced by emotional stimuli,”arXiv preprint arXiv:2307.11760, 2023

  20. [28]

    Automatic chain of thought prompting in large language models,

    Z. Zhang, A. Zhang, M. Li, and A. Smola, “Automatic chain of thought prompting in large language models,”preprint arXiv:2210.03493, 2022

  21. [29]

    Enhancing zero-shot chain-of-thought reasoning in large language models through logic,

    X. Zhao, M. Li, W. Lu, C. Weber, J. H. Lee, K. Chu, and S. Wermter, “Enhancing zero-shot chain-of-thought reasoning in large language models through logic,”arXiv preprint arXiv:2309.13339, 2023

  22. [30]

    Tree of thoughts: Deliberate problem solving with large language models,

    S. Yao, D. Yu, J. Zhao, I. Shafran, T. L. Griffiths, Y . Cao, and K. Narasimhan, “Tree of thoughts: Deliberate problem solving with large language models,”arXiv preprint arXiv:2305.10601, 2023

  23. [31]

    Beyond chain-of-thought, effec- tive graph-of-thought reasoning in language models,

    Y . Yao, Z. Li, and H. Zhao, “Beyond chain-of-thought, effec- tive graph-of-thought reasoning in language models,”arXiv preprint arXiv:2305.16582, 2023

  24. [32]

    React: Synergizing reasoning and acting in language models,

    S. Yao, J. Zhao, D. Yu, N. Du, I. Shafran, K. Narasimhan, and Y . Cao, “React: Synergizing reasoning and acting in language models,” in International Conference on Learning Representations (ICLR), 2023

  25. [33]

    Chain-of-knowledge: Grounding large language models via dynamic knowledge adapting over heterogeneous sources,

    X. Li, R. Zhao, Y . K. Chia, B. Ding, S. Joty, S. Poria, and L. Bing, “Chain-of-knowledge: Grounding large language models via dynamic knowledge adapting over heterogeneous sources,”arXiv preprint arXiv:2305.13269, 2023

  26. [34]

    Blip-2: Bootstrapping language- image pre-training with frozen image encoders and large language models,

    J. Li, D. Li, S. Savarese, and S. Hoi, “Blip-2: Bootstrapping language- image pre-training with frozen image encoders and large language models,”arXiv preprint arXiv:2301.12597, 2023

  27. [35]

    Flamingo: a visual language model for few-shot learning,

    J.-B. Alayrac, J. Donahue, P. Luc, A. Miech, I. Barr, Y . Hasson, K. Lenc, A. Mensch, K. Millican, M. Reynoldset al., “Flamingo: a visual language model for few-shot learning,”Advances in neural information processing systems, vol. 35, 2022

  28. [36]

    Language is not all you need: Aligning perception with language models,

    S. Huang, L. Dong, W. Wang, Y . Hao, S. Singhal, S. Ma, T. Lv, L. Cui, O. K. Mohammed, B. Patraet al., “Language is not all you need: Aligning perception with language models,”arXiv preprint arXiv:2302.14045, 2023

  29. [37]

    Kosmos-2: Grounding multimodal large language models to the world,

    Z. Peng, W. Wang, L. Dong, Y . Hao, S. Huang, S. Ma, and F. Wei, “Kosmos-2: Grounding multimodal large language models to the world,” arXiv preprint arXiv:2306.14824, 2023

  30. [38]

    Gpt4vis: what can gpt-4 do for zero-shot visual recognition?

    W. Wu, H. Yao, M. Zhang, Y . Song, W. Ouyang, and J. Wang, “Gpt4vis: what can gpt-4 do for zero-shot visual recognition?”arXiv preprint arXiv:2311.15732, 2023

  31. [39]

    Re- flexion: Language agents with verbal reinforcement learning,

    N. Shinn, F. Cassano, A. Gopinath, K. Narasimhan, and S. Yao, “Re- flexion: Language agents with verbal reinforcement learning,”Advances in Neural Information Processing Systems, vol. 36, 2023

  32. [40]

    Metacognitive prompting improves understand- ing in large language models,

    Y . Wang and Y . Zhao, “Metacognitive prompting improves understand- ing in large language models,”arXiv preprint arXiv:2308.05342, 2023

  33. [41]

    Self-explanation prompting improves dialogue understanding in large language models,

    H. Gao, T.-E. Lin, H. Li, M. Yang, Y . Wu, W. Ma, and Y . Li, “Self-explanation prompting improves dialogue understanding in large language models,”arXiv preprint arXiv:2309.12940, 2023

  34. [42]

    Take a step back: Evoking reasoning via abstraction in large language models,

    H. S. Zheng, S. Mishra, X. Chen, H.-T. Cheng, E. H. Chi, Q. V . Le, and D. Zhou, “Take a step back: Evoking reasoning via abstraction in large language models,”arXiv preprint arXiv:2310.06117, 2023

  35. [43]

    A critical assessment of the design issues in e-commerce systems development,

    S. A. Ehikioya and E. Guillemot, “A critical assessment of the design issues in e-commerce systems development,”Eng. Reports, vol. 2, 2020

  36. [44]

    Evidences of the mismatch be- tween industry and academy on modelling language quality evaluation,

    F. D. Giraldo, S. Espa ˜na, and ´O. Pastor, “Evidences of the mismatch be- tween industry and academy on modelling language quality evaluation,” ArXiv, vol. abs/1606.02025, 2016. APPENDIX [P1] Siddeshwar, Vaishali; Alwidian, Sanaa; Makrehchi, Masoud. A Comparative Study of Lar...

  37. [2025]

    Available: https://arxiv.org/abs/2502.18694

    [Online]. Available: https://arxiv.org/abs/2502.18694

Pith tools

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