Pith. sign in

REVIEW 4 major objections 5 minor 156 references

When LLMs Team Up: The Emergence of Collaborative Affective Computing

T0 review · 4 major / 5 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 The paper proposes the first systematic taxonomy of LLM-based collaboration for affective computing, grounded in the psychological split between fast intuitive and slow deliberative thinking, and claims that instance-level data…

desk verdict A useful taxonomy of LLM collaboration strategies for affective computing, but the empirical comparisons do not support the causal claims. read the letter →

arxiv 2506.01698 v1 pith:66JZ6YPQ submitted 2025-06-02 cs.CL cs.AI

Wenna Lai , Haoran Xie , Guandong Xu , Qing Li , S. Joe Qin This is my paper
classification cs.CLcs.AI
keywords affectivecomputinglargelanguagemodelscollaborativeAImulti-agentsystemsdualprocesstheoryaspect-basedsentimentanalysisemotionrecognitioninconversationemotionalsupport
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

Large language models are good at fluent language but struggle with the fast, intuitive grasp of emotion and the slow, deliberative reasoning that together make human affective intelligence. This paper argues that the fix is collaboration, and its contribution is a systematic map of how LLMs can team up with specialized models and with each other: structured collaboration, where a small model hands the LLM emotional signals at the token, instance, or modality level, and autonomous collaboration, where multiple LLM agents negotiate, vote, or role-play. To support the map, the paper runs its own comparisons on three representative affective computing tasks and claims that the winning strategy depends on the task: instance-level augmentation for aspect-based sentiment analysis, modal-level augmentation for emotion recognition in conversation, and scenario-immersive multi-agent simulation for emotional support conversation. The larger claim is that this taxonomy, grounded in the psychological distinction between fast and slow thinking, is the first systematic framework for LLM-based collaborative affective computing.

What carries the argument

The load-bearing machinery is the taxonomy itself, defined by the level at which information crosses between the specialized model and the LLM. Token-level augmentation injects signals into the token stream (retrieved demonstrations, external knowledge, or distilled reasoning chains); instance-level augmentation expands the training set with whole synthetic samples or auxiliary-task labels; modal-level augmentation either encodes non-text modalities into feature vectors or decodes them into textual affective cues; and autonomous collaboration lets multi-agent systems interact through consensus-driven debate or scenario-immersive role-play. The paper treats these categories as the explanatory variable that organizes both its literature review and the experimental comparisons in Tables 1–3.

What would settle it

Run a controlled experiment on ABSA that fixes the base model, assist model, prompt design, and training budget, varying only whether augmentation is token-level, instance-level, or multi-agent; if instance-level augmentation no longer outperforms the others, the paper's comparative conclusion fails.

Watch

Extended reading notes

Core claim

On its own terms, the paper's discovery is that LLM collaboration in affective computing is not one technique but a spectrum with a principled ordering. Structured collaboration keeps a fixed division of labor: one component supplies fast, intuitive affective perception (emotion categories, cause words, retrieved demonstrations, knowledge snippets, synthetic samples, or decoded multimodal cues) and a backbone LLM supplies slow, deliberative reasoning and generation. Autonomous collaboration removes the fixed roles: multiple LLM agents internalize both kinds of thinking and reach decisions through consensus-driven interaction or evolve states through scenario-immersive simulation. The paper claims these categories align with the psychological Dual Process Theory and the Affective Intelligence Theory, and it reports experiments indicating that the optimal collaboration level tracks the task: fine-grained aspect sentiment favors instance-level augmentation; conversational emotion recognition favors modal-level augmentation; and emotional support dialogue favors scenario-immersive multi-agent systems, especially when specialized models provide feedback.

Load-bearing premise

The paper assumes that the performance differences in Tables 1–3 are caused by the collaboration strategy category, even though the compared methods also differ in base model size, assist model, API costs, prompt design, fine-tuning setup, and training data, and none of these factors is controlled or statistically tested.

Editorial extensions

If this is right

  • If the taxonomy is right, new collaborative affective systems can be located on a single map, and their trade-offs between controllability, adaptation, compute, and API cost can be anticipated before deployment.
  • For aspect-based sentiment analysis, the paper's result says that investing in high-quality diverse training data through instance-level augmentation will beat both prompt-level knowledge injection and multi-agent debate.
  • For emotion recognition in conversation, the paper's result says text-only LLM pipelines are not enough; encoding or decoding audio and visual cues yields the largest gains.
  • For emotional support conversation, the paper's result says role-playing multi-agent systems with specialized feedback models can match or exceed fine-tuned single models at lower training cost.
  • Strategy choice should follow task type: structured augmentation suits precise understanding tasks, while autonomous simulation suits open-ended generation tasks.

Reading between the lines

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

  • A fair test of the taxonomy would hold base model, assist model, prompt design, and training budget fixed and vary only the collaboration mechanism; the current tables vary all of these jointly, so the reported rankings are open to alternative explanations.
  • The taxonomy suggests a testable prediction: the optimal collaboration level should track the affective granularity of the task—coarse sentiment may only need token-level signals, whereas situationally grounded emotion needs instance- or scenario-level interaction.
  • Nothing in the framework forbids nesting strategies, so hybrids such as multi-agent debate over instance-level augmented data are a natural untested extension.
  • If the Dual Process mapping is literal, then performance should degrade when the slow-thinking LLM is asked to do fast emotional perception or when the fast specialized model is asked to do slow reasoning; swapping roles while holding compute constant would test this.
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 / 5 minor

Summary. This survey organizes recent work on LLM-based collaboration for affective computing into a taxonomy of structured collaboration (token-level, instance-level, and modal-level augmentation) versus autonomous collaboration (multi-agent systems), motivated by Kahneman's Dual Process Theory and Affective Intelligence Theory. After reviewing methods, the paper reports new experiments on three tasks—ABSA, ERC, and ESC—comparing strategies from its taxonomy, and draws conclusions such as 'instance-level augmentation strategies significantly outperform token-level augmentation and multi-agent systems in ABSA' (§5.1.1). The paper also discusses challenges and future directions for collaborative affective AI.

Significance. The proposed taxonomy is a useful organizing device for a scattered and fast-growing literature, and the survey covers a broad set of recent methods with mechanism-level descriptions and formal notation. The authors are transparent about including their own prior methods (RVISA, MT-ISA) in the comparison. If the empirical claims were supported, this would be a valuable guide for practitioners choosing among collaboration strategies. However, the experimental comparisons in Tables 1–3 do not currently control for backbone model, assist model, training data size, or tuning protocol, and no uncertainty or significance measures are reported. The comparative conclusions in Section 5 therefore go beyond what the evidence supports. The taxonomy and survey content can stand, but the experimental argument needs substantial revision or reframing.

major comments (4)
  1. [§4.1.2, Table 1, §5.1.1] The claim that instance-level augmentation 'significantly outperforms' token-level augmentation and multi-agent systems in ABSA is not supported by the reported comparison. The methods differ simultaneously in inference model (Vicuna-13B, T5-Base 220M, BERT 110M, Flan-T5-XXL 11B, GPT-4o+DeepSeek-R1), assist model (Llama-2-7B, Mixtral-8x7B, Vicuna-13B+WizardLM-13B, GPT-3.5-turbo, GPT-4o), training data size (approximately 12K to 2M), and whether fine-tuning is used. No standard deviations, confidence intervals, or significance tests are reported, so the word 'significantly' is unjustified and the observed differences cannot be attributed to the collaboration strategy category.
  2. [§4.2.2, Table 2, §5.1.2] The ERC comparison is similarly confounded, and the modal-level conclusion is especially uncontrolled. MERITS-L uses audio input with GPT-3.5-turbo and Whisper-large-v3 as assistants and RoBERTa-large as the inference model, while DialogueLLM uses ERNIE Bot 260B to generate video captions and a LLaMA-2-7B LoRA model; CKERC and InstructERC use different assistants and prompts. The large IEMOCAP gap for MERITS-L (86.48 versus 72.40) may reflect the additional audio modality, the larger assistant, or other protocol differences, but the design does not isolate modality as the causal factor. Additionally, the missing EmoryNLP value for MERITS-L leaves the cross-dataset generalization claim incomplete.
  3. [§4.3.2, Table 3, §5.1.3] The ESC comparison is incomplete and internally inconsistent. Instance-level augmentation is explicitly excluded in §4.3.1, so the ESC section cannot support comparative statements about that strategy. The multi-agent systems use different inference models (GPT-3.5-turbo, BART 139M, BlenderBot-small 90M) and different assist models, and they are compared against token-level augmentation using yet another backbone (GPT-4o or Llama-3.1 LoRA). Many table cells are missing or unlabeled (e.g., BLEU-1/BLEU-3 for several rows, and all diversity metrics for VLESA), and no significance testing is reported. The conclusion that scenario-immersive multi-agent simulation shows the 'highest performance' holds only for a subset of automatic metrics and not across the full table.
  4. [§4.1.2] The interpretation that 'data quality over sheer model size' matters is post hoc. No experiment varies data quality while holding model family, model size, and training-data quantity fixed, nor does any ablation separate assist-model quality from augmentation-data quality. The sentence should be presented as a hypothesis suggested by the results, not as a conclusion established by the comparison.
minor comments (5)
  1. [Table 1] The table header reads 'F1 Score(marco)'; this should be 'macro-F1'.
  2. [§4.2.1] The text contains a duplicated period in 'sad..', and Table 3 uses stray hyphens for unavailable metrics, making it hard to distinguish missing values from zero values.
  3. [§3.2.1] The citation of 'Sun et al. [54]' for retrieval-augmented demonstration with RoBERTa-Large does not match reference [54], which is the LLM-negotiation paper; this citation needs correction or replacement.
  4. [§4.3.1] The exclusion of instance-level augmentation for fairness should be stated in the subsection's first paragraph, and Section 5.1.3 should explicitly acknowledge that the ESC comparison therefore omits one of the three structured-collaboration strategies.
  5. [Equations (1)–(9)] The notation reuses D, x, y, and M across strategies without per-section definitions; defining these symbols locally would improve readability.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the survey's comparative claims are empirical readings of externally benchmarked methods, and the authors' own methods are included transparently rather than used as load-bearing self-citations.

full rationale

This paper is a survey with a comparative experiment, not a derivation chain. Its taxonomy (structured vs. autonomous collaboration; token-, instance-, and modal-level augmentation; multi-agent systems) is a classification imposed on existing methods, and its conclusions in Sections 5.1.1-5.1.3 are empirical readings of Tables 1-3. No parameter is fitted and later renamed as a prediction, and no equation defines one strategy in terms of the outcome it is supposed to explain. The only self-citation cluster is RVISA [15] and MT-ISA [120] in Table 1, with BERT+CADA [116] also sharing co-authors, but these methods are evaluated on standard public benchmarks (SemEval-2014 Rest14 and Lap14) with externally checkable F1 scores, so the cited evidence does not reduce to the present paper's own assertions. The statistical weakness of the cross-table comparison, where backbone scale, assist model, API costs, tuning protocol, and training data differ across compared systems, is a validity and correctness concern rather than a circularity concern, because the strategy categories are not defined by those performance outcomes. No uniqueness theorem is imported from the authors, no ansatz is smuggled in via citation, and the Dual Process Theory grounding is an external psychological framework rather than a self-referential construction. The 'first systematic framework' claim is a novelty assertion, not a derivation. Overall, there is no significant circularity; the minor presence of the authors' own methods among the compared systems is transparent and not load-bearing, so the paper receives a score of 1.

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

The central claims rest on assumptions about the transferability of psychological theory to LLM systems, the representativeness of the selected methods, and the comparability of results across heterogeneous experimental setups. No free parameters or invented entities are introduced.

assumptions (3)
  • domain assumption Dual Process Theory and Affective Intelligence Theory adequately describe the cognitive division of labor in LLM collaboration systems.
    The entire taxonomy is motivated by Kahneman's System 1/System 2 analogy and Affective Intelligence Theory, but the paper provides no empirical validation that this psychological framing is a good model for LLM behavior. This enters in Section 1 and Section 3.1.
  • domain assumption The methods selected for experimental comparison are representative of the taxonomy categories.
    For each strategy type, only one or two existing methods are evaluated, and the paper assumes these are sufficient to draw conclusions about the whole category. This is assumed throughout Section 4.
  • domain assumption Results from different papers and different experimental setups are directly comparable.
    Tables 1-3 pool numbers from methods that use different assist models, base models, fine-tuning procedures, and cost budgets, yet the paper treats them as a level playing field for strategy comparison. This is the load-bearing assumption behind the findings in Section 5.1.

how reviews work

0 comments
Cite this review

Pith. "Pith review of When LLMs Team Up: The Emergence of Collaborative Affective Computing." pith.science (2026). https://pith.science/paper/66JZ6YPQ

@misc{pith2026250601698,
  author       = {Pith},
  title        = {Pith review of: When LLMs Team Up: The Emergence of Collaborative Affective Computing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/66JZ6YPQ}},
  note         = {Machine review of arXiv:2506.01698}
}
read the original abstract

Affective Computing (AC) is essential in bridging the gap between human emotional experiences and machine understanding. Traditionally, AC tasks in natural language processing (NLP) have been approached through pipeline architectures, which often suffer from structure rigidity that leads to inefficiencies and limited adaptability. The advent of Large Language Models (LLMs) has revolutionized this field by offering a unified approach to affective understanding and generation tasks, enhancing the potential for dynamic, real-time interactions. However, LLMs face cognitive limitations in affective reasoning, such as misinterpreting cultural nuances or contextual emotions, and hallucination problems in decision-making. To address these challenges, recent research advocates for LLM-based collaboration systems that emphasize interactions among specialized models and LLMs, mimicking human-like affective intelligence through the synergy of emotional and rational thinking that aligns with Dual Process Theory in psychology. This survey aims to provide a comprehensive overview of LLM-based collaboration systems in AC, exploring from structured collaborations to autonomous collaborations. Specifically, it includes: (1) A systematic review of existing methods, focusing on collaboration strategies, mechanisms, key functions, and applications; (2) Experimental comparisons of collaboration strategies across representative tasks in affective understanding and generation; (3) An analysis highlighting the potential of these systems to enhance robustness and adaptability in complex affective reasoning; (4) A discussion of key challenges and future research directions to further advance the field. This work is the first to systematically explore collaborative intelligence with LLMs in AC, paving the way for more powerful applications that approach human-like social intelligence.

Figures

Figures reproduced from arXiv: 2506.01698 by the authors.

Figure 1
Figure 1. These systems handle inputs through sequential [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 1
Figure 1. An overview of the LLM-based collaboration system compared to the conventional deep learning system [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Development of LLM-based collaboration, regarding the intensity of system 1 (intuitive thinking) and system 2 [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figures from the paper (5 more)
Figure 3
Figure 3. Figure 3: Taxonomy of LLM-based collaboration strategies with related references, organized separately for the affective [PITH_FULL_IMAGE:figures/full_fig_p006_3.png]
Figure 4
Figure 4. Figure 4: Token-level augmentation strategies in structured collaboration. Three strategies are illustrated with examples, [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Instance-level augmentation strategies in structured collaboration. Two strategies are illustrated with examples, [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Modal-level augmentation strategies in structured collaboration. Two strategies are illustrated with examples, [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Multi-agent systems for autonomous collaboration. Two strategies are illustrated with examples, including [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

156 extracted references · 75 canonical work pages

  1. [1]

    R. W. Picard, Affective computing. MIT press, 1997

    1997

  2. [2]

    Toward machine emotional intelligence: Analysis of affective physiological state,

    R. W. Picard, E. Vyzas, and J. Healey, “Toward machine emotional intelligence: Analysis of affective physiological state,”IEEE Trans. Pattern Anal. Mach. Intell., vol. 23, no. 10, pp. 1175–1191, 2001

    2001

  3. [3]

    A circumplex model of affect

    J. A. Russell, “A circumplex model of affect.” Journal of Personality and Social Psychology, vol. 39, pp. 1161–1178, 1980

    1980

  4. [4]

    Defining affect in relation to cognition: A re- sponse to susan mcleod,

    K. S. Fleckenstein, “Defining affect in relation to cognition: A re- sponse to susan mcleod,” Journal of Advanced Composition, vol. 11, no. 2, pp. 447–453, 1991

    1991

  5. [5]

    Editorial: The social nature of emotions,

    G. A. van Kleef, A. Cheshin, A. H. Fischer, and I. K. Schneider, “Editorial: The social nature of emotions,” Frontiers in Psychology, vol. 7, 2016. [Online]. Available: https://api.semanticscholar. org/CorpusID:5209023

    2016

  6. [6]

    Affective computing and sentiment analysis,

    E. Cambria, “Affective computing and sentiment analysis,” IEEE Intell. Syst., vol. 31, no. 2, p. 102–107, Mar. 2016

    2016

  7. [7]

    A systematic review on affective computing: emotion models, databases, and recent advances,

    Y. Wang, W. Song, W. Tao, A. Liotta, D. Yang, X. Li, S. Gao, Y. Sun, W. Ge, W. Zhang, and W. Zhang, “A systematic review on affective computing: emotion models, databases, and recent advances,” Inf. Fusion, vol. 83-84, pp. 19–52, 2022

    2022

  8. [8]

    A comprehensive survey on affective computing: Challenges, trends, applications, and future directions,

    S. Afzal, H. A. Khan, M. J. Piran, and J. Lee, “A comprehensive survey on affective computing: Challenges, trends, applications, and future directions,” IEEE Access , vol. 12, pp. 96 150–96 168, 2024

    2024

Show all 156 references
  1. [9]

    Affective computing: A review,

    J. Tao and T. Tan, “Affective computing: A review,” in ACII, ser. Lecture Notes in Computer Science, vol. 3784. Springer, 2005, pp. 981–995

    2005

  2. [10]

    A survey of autonomous human affect detection methods for social robots engaged in natural hri,

    D. McColl, A. Hong, N. Hatakeyama, G. Nejat, and B. Benhabib, “A survey of autonomous human affect detection methods for social robots engaged in natural hri,” Journal of Intelligent & Robotic Systems, vol. 82, pp. 101 – 133, 2015

    2015

  3. [11]

    The affect dilemma for artificial agents: Should we develop affective artificial agents?

    M. Scheutz, “The affect dilemma for artificial agents: Should we develop affective artificial agents?” IEEE Trans. Affect. Comput. , vol. 3, no. 4, pp. 424–433, 2012

    2012

  4. [12]

    Affective computing in the era of large language models: A survey from the nlp perspective,

    Y. Zhang, X. Yang, X. Xu, Z. Gao, Y. Huang, S. Mu, S. Feng, D. Wang, Y. Zhang, K. Song, and G. Yu, “Affective computing in the era of large language models: A survey from the nlp perspective,” arXiv, 2024

    2024

  5. [13]

    Is chatgpt equipped with emotional dialogue capabilities?

    W. Zhao, Y. Zhao, X. Lu, S. Wang, Y. Tong, and B. Qin, “Is chatgpt equipped with emotional dialogue capabilities?” arXiv, 2023

    2023

  6. [14]

    Affective computing: Recent advances, challenges, and future trends,

    G. Pei, H. Li, Y. Lu, Y. Wang, S. Hua, and T. Li, “Affective computing: Recent advances, challenges, and future trends,” Intelligent Computing, vol. 3, p. 0076, 2024

    2024

  7. [15]

    Rvisa: Reasoning and verifica- tion for implicit sentiment analysis,

    W. Lai, H. Xie, G. Xu, and Q. Li, “Rvisa: Reasoning and verifica- tion for implicit sentiment analysis,” IEEE Trans. Affect. Comput., pp. 1–12, 2025

    2025

  8. [16]

    Sentiment analysis in the era of large language models: A reality check,

    W. Zhang, Y. Deng, B. Liu, S. Pan, and L. Bing, “Sentiment analysis in the era of large language models: A reality check,” in Findings of NAACL, 2024, pp. 3881–3906

    2024

  9. [17]

    A wide evaluation of chatgpt on affective computing tasks,

    M. M. Amin, R. Mao, E. Cambria, and B. W. Schuller, “A wide evaluation of chatgpt on affective computing tasks,” IEEE Trans. Affect. Comput., vol. 15, no. 4, pp. 2204–2212, 2024

    2024

  10. [18]

    New approach in quantification of emotional intensity from the speech signal: emotional temperature,

    J. B. Alonso, J. Cabrera, M. Medina, and C. M. Travieso, “New approach in quantification of emotional intensity from the speech signal: emotional temperature,” Expert Syst. Appl., vol. 42, no. 24, pp. 9554–9564, 2015

    2015

  11. [19]

    CEM: commonsense-aware empathetic response generation,

    S. Sabour, C. Zheng, and M. Huang, “CEM: commonsense-aware empathetic response generation,” in AAAI, 2022, pp. 11 229– 11 237

    2022

  12. [20]

    Emotion-aware and intent-controlled empathetic response generation using hierarchical transformer network,

    T. Saha and S. Ananiadou, “Emotion-aware and intent-controlled empathetic response generation using hierarchical transformer network,” in IJCNN, 2022, pp. 1–8

    2022

  13. [21]

    Sequential models for sentiment analysis: A comparative study,

    O. O. Adebanji, I. Gelbukh, H. Calvo, and O. E. Ojo, “Sequential models for sentiment analysis: A comparative study,” in MICAI. Springer Nature Switzerland, 2022, pp. 227–235

    2022

  14. [22]

    Empmff: A multi-factor sequence fusion framework for empathetic re- sponse generation,

    X. Pang, Y. Wang, S. Fan, L. Chen, S. Shang, and P . Han, “Empmff: A multi-factor sequence fusion framework for empathetic re- sponse generation,” in Proceedings of the ACM Web Conference 2023, 2023, p. 1754–1764

    2023

  15. [23]

    Control globally, understand locally: A global-to-local hierarchical graph network for emotional support conversation,

    W. Peng, Y. Hu, L. Xing, Y. Xie, Y. Sun, and Y. Li, “Control globally, understand locally: A global-to-local hierarchical graph network for emotional support conversation,” in IJCAI, 2022, pp. 4324–4330

    2022

  16. [24]

    Towards empathetic open-domain conversation models: A new benchmark and dataset,

    H. Rashkin, E. M. Smith, M. Li, and Y. Boureau, “Towards empathetic open-domain conversation models: A new benchmark and dataset,” in Proceedings of the 57th Conference of the Association for Computational Linguistics, ACL 2019, Florence, Italy, July 28- August 2, 2019, Volume...

    2019 doi

  17. [25]

    Empathetic dialogue generation via sensitive emotion recogni- tion and sensible knowledge selection,

    L. Wang, J. Li, Z. Lin, F. Meng, C. Yang, W. Wang, and J. Zhou, “Empathetic dialogue generation via sensitive emotion recogni- tion and sensible knowledge selection,” in Findings of EMNLP , 2022, pp. 4634–4645

    2022

  18. [26]

    Large language models are zero-shot reasoners,

    T. Kojima, S. S. Gu, M. Reid, Y. Matsuo, and Y. Iwasawa, “Large language models are zero-shot reasoners,” in NeurIPS, 2022. 16 When LLMs Team Up: The Emergence of Collaborative Affective Computing

    2022

  19. [27]

    Finetuned language models are zero- shot learners,

    J. Wei, M. Bosma, V . Y. Zhao, K. Guu, A. W. Yu, B. Lester, N. Du, A. M. Dai, and Q. V . Le, “Finetuned language models are zero- shot learners,” in ICLR, 2022

    2022

  20. [28]

    Unifiedabsa: A unified absa frame- work based on multi-task instruction tuning,

    Z. Wang, R. Xia, and J. Yu, “Unifiedabsa: A unified absa frame- work based on multi-task instruction tuning,” arXiv, 2022

    2022

  21. [29]

    UniMSE: Towards unified multimodal sentiment analysis and emotion recognition,

    G. Hu, T.-E. Lin, Y. Zhao, G. Lu, Y. Wu, and Y. Li, “UniMSE: Towards unified multimodal sentiment analysis and emotion recognition,” in EMNLP, Dec. 2022

    2022

  22. [30]

    LEGO-ABSA: A prompt-based task assemblable unified genera- tive framework for multi-task aspect-based sentiment analysis,

    T. Gao, J. Fang, H. Liu, Z. Liu, C. Liu, P . Liu, Y. Bao, and W. Yan, “LEGO-ABSA: A prompt-based task assemblable unified genera- tive framework for multi-task aspect-based sentiment analysis,” in COLING, 2022

    2022

  23. [31]

    Self-refine: Iterative refinement with self-feedback,

    A. Madaan, N. Tandon, P . Gupta, S. Hallinan, L. Gao, S. Wiegr- effe, U. Alon, N. Dziri, S. Prabhumoye, Y. Yang, S. Gupta, B. P . Majumder, K. Hermann, S. Welleck, A. Yazdanbakhsh, and P . Clark, “Self-refine: Iterative refinement with self-feedback,” in NeurIPS, 2023

    2023

  24. [32]

    When can llms Actually correct their own mistakes? A critical survey of self- correction of llms,

    R. Kamoi, Y. Zhang, N. Zhang, J. Han, and R. Zhang, “When can llms Actually correct their own mistakes? A critical survey of self- correction of llms,” Trans. Assoc. Comput. Linguistics , vol. 12, pp. 1417–1440, 2024

    2024

  25. [33]

    Can large language model agents simulate human trust behavior?

    C. Xie, C. Chen, F. Jia, Z. Ye, S. Lai, K. Shu, J. Gu, A. Bibi, Z. Hu, D. Jurgens, J. Evans, P . Torr, B. Ghanem, and G. Li, “Can large language model agents simulate human trust behavior?” in NeurIPS, 2024

    2024

  26. [34]

    Multi-agent collaboration mechanisms: A survey of llms,

    K.-T. Tran, D. Dao, M.-D. Nguyen, Q.-V . Pham, B. O’Sullivan, and H. D. Nguyen, “Multi-agent collaboration mechanisms: A survey of llms,” arXiv, 2025

    2025

  27. [35]

    Towards measuring and modeling

    M. F. Adilazuarda, S. Mukherjee, P . Lavania, S. Singh, A. F. Aji, J. O’Neill, A. Modi, and M. Choudhury, “Towards measuring and modeling "culture" in llms: A survey,” in EMNLP, 2024, pp. 15 763–15 784

    2024

  28. [36]

    Culturellm: Incorporating cultural differences into large language models,

    C. Li, M. Chen, J. Wang, S. Sitaram, and X. Xie, “Culturellm: Incorporating cultural differences into large language models,” in NeurIPS, 2024

    2024

  29. [37]

    Siren’s song in the ai ocean: A survey on hallucination in large language models,

    Y. Zhang, Y. Li, L. Cui, D. Cai, L. Liu, T. Fu, X. Huang, E. Zhao, Y. Zhang, Y. Chen, L. Wang, A. T. Luu, W. Bi, F. Shi, and S. Shi, “Siren’s song in the ai ocean: A survey on hallucination in large language models,” arXiv, 2023

    2023

  30. [38]

    A survey on the honesty of large language models,

    S. Li, C. Yang, T. Wu, C. Shi, Y. Zhang, X. Zhu, Z. Cheng, D. Cai, M. Yu, L. Liu, J. Zhou, Y. Yang, N. Wong, X. Wu, and W. Lam, “A survey on the honesty of large language models,” arXiv, 2024

    2024

  31. [39]

    Kahneman, Thinking, fast and slow

    D. Kahneman, Thinking, fast and slow. Straus and Giroux, 2011

    2011

  32. [40]

    Collaborative intelligence: A scoping review of current appli- cations,

    E. Schleiger, C. Mason, C. Naughtin, A. Reeson, and C. P . and, “Collaborative intelligence: A scoping review of current appli- cations,” Applied Artificial Intelligence , vol. 38, no. 1, p. 2327890, 2024

    2024

  33. [41]

    Ai and the future of collaborative work: Group ideation with an llm in a virtual canvas,

    J. He, S. Houde, G. E. Gonzalez, D. A. Silva Moran, S. I. Ross, M. Muller, and J. D. Weisz, “Ai and the future of collaborative work: Group ideation with an llm in a virtual canvas,” in Pro- ceedings of the 3rd Annual Meeting of the Symposium on Human- Computer Interaction for...

    2024

  34. [42]

    Collaborative ai in sentiment analysis: System architecture, data prediction and deployment strategies,

    C. Zhang, J. Hou, X. Tan, G. Li, and C. Chen, “Collaborative ai in sentiment analysis: System architecture, data prediction and deployment strategies,” arXiv, 2024

    2024

  35. [43]

    What is the role of small models in the llm era: A survey,

    L. Chen and G. Varoquaux, “What is the role of small models in the llm era: A survey,” arXiv, 2025

    2025

  36. [44]

    A comprehensive survey of small language models in the era of large language models: Techniques, enhancements, applications, collaboration with llms, and trustworthiness,

    F. Wang, Z. Zhang, X. Zhang, Z. Wu, T. Mo, Q. Lu, W. Wang, R. Li, J. Xu, X. Tang, Q. He, Y. Ma, M. Huang, and S. Wang, “A comprehensive survey of small language models in the era of large language models: Techniques, enhancements, applications, collaboration with llms, and tru...

    2024

  37. [45]

    Marcus, W

    G. Marcus, W. Neuman, and M. MacKuen, Affective Intelligence and Political Judgment , ser. Affective Intelligence and Political Judgment. University of Chicago Press, 2000

    2000

  38. [46]

    Enhancing empathetic response generation by augmenting llms with small-scale empathetic models,

    Z. Yang, Z. Ren, W. Yufeng, S. Peng, H. Sun, X. Zhu, and X. Liao, “Enhancing empathetic response generation by augmenting llms with small-scale empathetic models,” arXiv, 2024

    2024

  39. [47]

    Artificial social intelligence in teamwork: How team traits influence human-ai dynamics in complex tasks

    B. Rhyse, J. Williams, S. M. Fiore, and F. Jentsch, “Artificial social intelligence in teamwork: How team traits influence human-ai dynamics in complex tasks.” Frontiers in Robotics and AI , vol. 12, 2025

    2025

  40. [48]

    A survey on llm-based multi-agent systems: workflow, infrastructure, and challenges,

    X. Li, S. Wang, S. Zeng, Y. Wu, and Y. Yang, “A survey on llm-based multi-agent systems: workflow, infrastructure, and challenges,” Vicinagearth, 2024

    2024

  41. [49]

    Large language model based multi- agents: A survey of progress and challenges,

    T. Guo, X. Chen, Y. Wang, R. Chang, S. Pei, N. V . Chawla, O. Wiest, and X. Zhang, “Large language model based multi- agents: A survey of progress and challenges,” in IJCAI, 2024, pp. 8048–8057

    2024

  42. [50]

    Can small language models help large language models reason better?: LM-guided chain-of-thought,

    J. Lee, F. Yang, T. Tran, Q. Hu, E. Barut, and K.-W. Chang, “Can small language models help large language models reason better?: LM-guided chain-of-thought,” in LREC-COLING, Torino, Italia, May 2024, pp. 2835–2843

    2024

  43. [51]

    Survey on sentiment analysis: evolution of research methods and topics,

    J. Cui, Z. Wang, S.-B. Ho, and E. Cambria, “Survey on sentiment analysis: evolution of research methods and topics,” Artif. Intell. Rev., 2023

    2023

  44. [52]

    Multimodal sentiment analysis: A survey of methods, trends, and challenges,

    R. Das and T. D. Singh, “Multimodal sentiment analysis: A survey of methods, trends, and challenges,” ACM Comput. Surv., 2023

    2023

  45. [53]

    A survey on aspect-based sentiment analysis: Tasks, methods, and chal- lenges,

    W. Zhang, X. Li, Y. Deng, L. Bing, and W. Lam, “A survey on aspect-based sentiment analysis: Tasks, methods, and chal- lenges,” IEEE Trans. Knowl. Data Eng., 2022

    2022

  46. [54]

    Sentiment analysis through llm negotiations,

    X. Sun, X. Li, S. Zhang, S. Wang, F. Wu, J. Li, T. Zhang, and G. Wang, “Sentiment analysis through llm negotiations,” arXiv, 2023

    2023

  47. [55]

    Large language models performance comparison of emotion and sentiment classification,

    W. Stigall, M. A. A. H. Khan, D. C. Attota, F. Nweke, and Y. Pei, “Large language models performance comparison of emotion and sentiment classification,” in Proceedings of the 2024 ACM Southeast Conference, 2024, pp. 60–68

    2024

  48. [56]

    Mvp: Multi-view prompting improves aspect sentiment tuple prediction,

    Z. Gou, Q. Guo, and Y. Yang, “Mvp: Multi-view prompting improves aspect sentiment tuple prediction,” in ACL, 2023

    2023

  49. [57]

    Star: Stepwise task augmenta- tion and relation learning for aspect sentiment quad prediction,

    W. Lai, H. Xie, G. Xu, and Q. Li, “Star: Stepwise task augmenta- tion and relation learning for aspect sentiment quad prediction,” arXiv, 2025

    2025

  50. [58]

    Instructabsa: Instruction learning for aspect based sentiment analysis,

    K. Scaria, H. Gupta, S. Goyal, S. A. Sawant, S. Mishra, and C. Baral, “Instructabsa: Instruction learning for aspect based sentiment analysis,” in NAACL, 2024

    2024

  51. [59]

    Aspect-based senti- ment analysis using bert,

    M. Hoang, O. A. Bihorac, and J. Rouces, “Aspect-based senti- ment analysis using bert,” in Nordic Conference of Computational Linguistics, 2019

    2019

  52. [60]

    WRIME: A new dataset for emotional intensity estimation with subjective and objective annotations,

    T. Kajiwara, C. Chu, N. Takemura, Y. Nakashima, and H. Naga- hara, “WRIME: A new dataset for emotional intensity estimation with subjective and objective annotations,” in NAACL-HLT. As- sociation for Computational Linguistics, 2021, pp. 2095–2104

    2021

  53. [61]

    Improving representation with hier- archical contrastive learning for emotion-cause pair extraction,

    G. Hu, Y. Zhao, and G. Lu, “Improving representation with hier- archical contrastive learning for emotion-cause pair extraction,” IEEE Trans. Affect. Comput., vol. 15, no. 4, pp. 1997–2011, 2024

    1997

  54. [62]

    A reinforcement learning based two-stage model for emotion cause pair extraction,

    X. Chen, Q. Li, Z. Li, H. Xie, F. L. Wang, and J. Wang, “A reinforcement learning based two-stage model for emotion cause pair extraction,” IEEE Trans. Affect. Comput. , vol. 14, no. 3, pp. 1779–1790, 2023. [Online]. Available: https://doi.org/10.1109/ TAFFC.2022.3218648

    2023

  55. [63]

    Emoprompt-ecpe: Emotion knowledge-aware prompt- tuning for emotion-cause pair extraction,

    X. Gu, Z. Zhou, Z. Meng, J. Li, T. Gomes, A. Tavares, and H. Xu, “Emoprompt-ecpe: Emotion knowledge-aware prompt- tuning for emotion-cause pair extraction,” in LREC/COLING, 2024, pp. 5678–5688

    2024

  56. [64]

    Liu, Sentiment Analysis - Mining Opinions, Sentiments, and Emotions

    B. Liu, Sentiment Analysis - Mining Opinions, Sentiments, and Emotions. Cambridge University Press, 2015

    2015

  57. [65]

    Beneath the tip of the iceberg: Current challenges and new directions in sentiment analysis research,

    S. Poria, D. Hazarika, N. Majumder, and R. Mihalcea, “Beneath the tip of the iceberg: Current challenges and new directions in sentiment analysis research,” IEEE Transactions on Affective Computing, vol. 14, pp. 108–132, 2020

    2020

  58. [66]

    Sarcasm detection using news headlines dataset,

    R. Misra and P . Arora, “Sarcasm detection using news headlines dataset,” AI Open, 2023

    2023

  59. [67]

    Is sarcasm detection a step- by-step reasoning process in large language models?

    B. Yao, Y. Zhang, Q. Li, and J. Qin, “Is sarcasm detection a step- by-step reasoning process in large language models?” in AAAI, 2025

    2025

  60. [68]

    Sentiment analysis and sarcasm detection using deep multi-task learning: Sentiment analysis and sarcasm detection,

    Y. Y. Tan, C.-O. Chow, J. Kanesan, J. H. Chuah, and Y. Lim, “Sentiment analysis and sarcasm detection using deep multi-task learning: Sentiment analysis and sarcasm detection,” Wirel. Pers. Commun., vol. 129, no. 3, p. 2213–2237, Mar. 2023

    2023

  61. [69]

    Topic- oriented sarcasm detection via entity knowledge-based prompt learning,

    Y. Zhou, S. Zhang, C. Wang, Y. Wang, X. Wang, and K. Li, “Topic- oriented sarcasm detection via entity knowledge-based prompt learning,” Comput. Sci. Inf. Syst. , vol. 22, no. 1, pp. 33–57, 2025. [Online]. Available: https://doi.org/10.2298/csis240320067z

    2025 doi

  62. [70]

    Instructerc: Reforming emotion recognition in conversation with multi-task retrieval-augmented large language models,

    S. Lei, G. Dong, X. Wang, K. Wang, R. Qiao, and S. Wang, “Instructerc: Reforming emotion recognition in conversation with multi-task retrieval-augmented large language models,” arXiv, 2024

    2024

  63. [71]

    Zhang, M

    Y. Zhang, M. Wang, Y. Wu, P . Tiwari, Q. Li, B. Wang, and J. Qin, “Dialoguellm: Context and emotion knowledge-tuned 17 When LLMs Team Up: The Emergence of Collaborative Affective Computing large language models for emotion recognition in conversations,” arXiv, 2024

    2024

  64. [72]

    Language models are few-shot learners,

    T. B. Brown, B. Mann, N. Ryder, M. Subbiah, J. Kaplan, P . Dhari- wal, A. Neelakantan, P . Shyam, G. Sastry, A. Askell, S. Agarwal, A. Herbert-Voss, G. Krueger, T. Henighan, R. Child, A. Ramesh, D. M. Ziegler, J. Wu, C. Winter, C. Hesse, M. Chen, E. Sigler, M. Litwin, S. Gray,...

    2020

  65. [73]

    Self-consistency improves chain of thought reasoning in language models,

    X. Wang, J. Wei, D. Schuurmans, Q. V . Le, E. H. Chi, S. Narang, A. Chowdhery, and D. Zhou, “Self-consistency improves chain of thought reasoning in language models,” in ICLR, 2023

    2023

  66. [74]

    Complexity- based prompting for multi-step reasoning,

    Y. Fu, H. Peng, A. Sabharwal, P . Clark, and T. Khot, “Complexity- based prompting for multi-step reasoning,” in ICLR, 2023

    2023

  67. [75]

    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,” in ICLR, 2023

    2023

  68. [76]

    Prompted opinion sum- marization with GPT-3.5,

    A. Bhaskar, A. Fabbri, and G. Durrett, “Prompted opinion sum- marization with GPT-3.5,” in Findings of ACL, Jul. 2023, pp. 9282– 9300

    2023

  69. [77]

    An iterative associative memory model for empathetic response generation,

    Z. Yang, Z. Ren, Y. Wang, H. Sun, C. Chen, X. Zhu, and X. Liao, “An iterative associative memory model for empathetic response generation,” in ACL, 2024, pp. 3081–3092

    2024

  70. [78]

    EmpCRL: Controllable empathetic response generation via in-context commonsense rea- soning and reinforcement learning,

    M. Cai, D. Wang, S. Feng, and Y. Zhang, “EmpCRL: Controllable empathetic response generation via in-context commonsense rea- soning and reinforcement learning,” in LREC-COLING, 2024, pp. 5734–5746

    2024

  71. [79]

    CASE: aligning coarse-to-fine cognition and affection for empathetic response generation,

    J. Zhou, C. Zheng, B. Wang, Z. Zhang, and M. Huang, “CASE: aligning coarse-to-fine cognition and affection for empathetic response generation,” in ACL, 2023, pp. 8223–8237

    2023

  72. [80]

    Control globally, understand locally: A global-to-local hierarchical graph network for emotional support conversation,

    W. Peng, Y. Hu, L. Xing, Y. Xie, Y. Sun, and Y. Li, “Control globally, understand locally: A global-to-local hierarchical graph network for emotional support conversation,” in IJCAI, 2022

    2022

  73. [81]

    Augesc: Dialogue augmentation with large language models for emo- tional support conversation,

    C. Zheng, S. Sabour, J. Wen, Z. Zhang, and M. Huang, “Augesc: Dialogue augmentation with large language models for emo- tional support conversation,” in ACL, 2022

    2022

  74. [82]

    Towards emotional support dialog systems,

    S. Liu, C. Zheng, O. Demasi, S. Sabour, Y. Li, Z. Yu, Y. Jiang, and M. Huang, “Towards emotional support dialog systems,” inACL, 2021

    2021

  75. [83]

    Towards mul- timodal emotional support conversation systems,

    Y. Chu, L. Liao, Z. Zhou, C.-W. Ngo, and R. Hong, “Towards mul- timodal emotional support conversation systems,” arXiv, 2024

    2024

  76. [84]

    Knowledge-enhanced mixed-initiative dialogue system for emotional support conver- sations,

    Y. Deng, W. Zhang, Y. Yuan, and W. Lam, “Knowledge-enhanced mixed-initiative dialogue system for emotional support conver- sations,” in ACL, 2023

    2023

  77. [85]

    Pal: Persona-augmented emotional support conversation genera- tion,

    J. Cheng, S. Sabour, H. Sun, Z. Chen, and M. Huang, “Pal: Persona-augmented emotional support conversation genera- tion,” in ACL, 2022

    2022

  78. [86]

    Iteratively calibrat- ing prompts for unsupervised diverse opinion summarization,

    J. Wang, Y. Sun, Y. Liang, X. Li, and B. Gong, “Iteratively calibrat- ing prompts for unsupervised diverse opinion summarization,” in ECAI, 2024

    2024

  79. [87]

    Hierarchical indexing for retrieval-augmented opinion summarization,

    T. Hosking, H. Tang, and M. Lapata, “Hierarchical indexing for retrieval-augmented opinion summarization,” Transactions of the Association for Computational Linguistics , vol. 12, pp. 1533–1555, 2024

    2024

  80. [88]

    Product description and qa assisted self-supervised opinion summarization,

    T. Siledar, R. Rangaraju, S. S. R. R. Muddu, S. Banerjee, A. Patil, S. Singh, M. Chelliah, N. Garera, S. Nath, and P . Bhattacharyya, “Product description and qa assisted self-supervised opinion summarization,” in NAACL-HLT, 2024

    2024

  81. [89]

    Improving multi-turn emotional support dialogue generation with lookahead strategy planning,

    Y. Cheng, W. Liu, W. Li, J. Wang, R. Zhao, B. Liu, X. Liang, and Y. Zheng, “Improving multi-turn emotional support dialogue generation with lookahead strategy planning,” in EMNLP, Dec. 2022, pp. 3014–3026

    2022

  82. [90]

    Emotionally intelligent chatbots: A systematic literature review,

    G. Bilquise, S. Ibrahim, and K. Shaalan, “Emotionally intelligent chatbots: A systematic literature review,” Human Behavior and Emerging Technologies, 2022

    2022

  83. [91]

    Self-chats from large language models make small emotional support chatbot better,

    Z. Zheng, L. Liao, Y. Deng, L. Qin, and L. Nie, “Self-chats from large language models make small emotional support chatbot better,” in Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , Aug. 2024, pp. 11 325–11 345

    2024

  84. [92]

    Can large language models be good emotional supporter? mitigating preference bias on emotional support conversation,

    D. Kang, S. Kim, T. Kwon, S. Moon, H. Cho, Y. Yu, D. Lee, and J. Yeo, “Can large language models be good emotional supporter? mitigating preference bias on emotional support conversation,” in ACL, 2024, pp. 15 232–15 261

    2024

  85. [93]

    Llama: Open and efficient foundation language models,

    H. Touvron, T. Lavril, G. Izacard, X. Martinet, M.-A. Lachaux, T. Lacroix, B. Rozière, N. Goyal, E. Hambro, F. Azhar, A. Ro- driguez, A. Joulin, E. Grave, and G. Lample, “Llama: Open and efficient foundation language models,” arXiv, 2023

    2023

  86. [94]

    ESC-eval: Evaluating emotion support conversations in large language models,

    H. Zhao, L. Li, S. Chen, S. Kong, J. Wang, K. Huang, T. Gu, Y. Wang, J. Wang, L. Dandan, Z. Li, Y. Teng, Y. Xiao, and Y. Wang, “ESC-eval: Evaluating emotion support conversations in large language models,” in Proceedings of the 2024 Conference on Empirical Methods in Natural L...

    2024

  87. [95]

    Make your LLM fully utilize the context,

    S. An, Z. Ma, Z. Lin, N. Zheng, J. Lou, and W. Chen, “Make your LLM fully utilize the context,” in NeurIPS, 2024

    2024

  88. [96]

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

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

    2022

  89. [97]

    Ds 2-absa: Dual-stream data synthesis with label refinement for few-shot aspect-based sentiment analysis,

    H. Xu, Y. Zhang, Q. Wang, and R. Xu, “Ds 2-absa: Dual-stream data synthesis with label refinement for few-shot aspect-based sentiment analysis,” arXiv, 2024

    2024

  90. [98]

    Two heads are better than one: Zero-shot cognitive reasoning via multi-llm knowledge fusion,

    L. Liu, D. Zhang, S. Li, G. Zhou, and E. Cambria, “Two heads are better than one: Zero-shot cognitive reasoning via multi-llm knowledge fusion,” in CIKM, 2024, pp. 1462–1472

    2024

  91. [99]

    Attention is all you need,

    A. Vaswani, N. M. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, “Attention is all you need,” in NeurIPS, 2017

    2017

  92. [100]

    Bert: Pre- training of deep bidirectional transformers for language under- standing,

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “Bert: Pre- training of deep bidirectional transformers for language under- standing,” in NAACL, 2019

    2019

  93. [101]

    Training language models to follow instructions with human feedback,

    L. Ouyang, J. Wu, X. Jiang, D. Almeida, C. L. Wainwright, P . Mishkin, C. Zhang, S. Agarwal, K. Slama, A. Ray, J. Schulman, J. Hilton, F. Kelton, L. Miller, M. Simens, A. Askell, P . Welinder, P . F. Christiano, J. Leike, and R. Lowe, “Training language models to follow instru...

    2022

  94. [102]

    Scaling instruction- finetuned language models,

    H. W. Chung, L. Hou, S. Longpre, B. Zoph, Y. Tay, W. Fedus, Y. Li, X. Wang, M. Dehghani, S. Brahma, A. Webson, S. S. Gu, Z. Dai, M. Suzgun, X. Chen, A. Chowdhery, A. Castro-Ros, M. Pellat, K. Robinson, D. Valter, S. Narang, G. Mishra, A. Yu, V . Zhao, Y. Huang, A. Dai, H. Yu, ...

    2022

  95. [103]

    Large language models are reasoning teachers,

    N. Ho, L. Schmid, and S. Yun, “Large language models are reasoning teachers,” in ACL (Volume 1: Long Papers), 2023

    2023

  96. [104]

    Distilling step-by-step! outperforming larger language models with less training data and smaller model sizes,

    C. Hsieh, C. Li, C. Yeh, H. Nakhost, Y. Fujii, A. Ratner, R. Krishna, C. Lee, and T. Pfister, “Distilling step-by-step! outperforming larger language models with less training data and smaller model sizes,” in Findings of ACL, 2023, pp. 8003–8017

    2023

  97. [105]

    Explanations from large language models make small reasoners better,

    S. Li, J. Chen, yelong shen, Z. Chen, X. Zhang, Z. Li, H. Wang, J. Qian, B. Peng, Y. Mao, W. Chen, and X. Yan, “Explanations from large language models make small reasoners better,” inSAI- AAAI, 2024

    2024

  98. [106]

    Lora: Low-rank adaptation of large language models,

    E. J. Hu, Y. Shen, P . Wallis, Z. Allen-Zhu, Y. Li, S. Wang, L. Wang, and W. Chen, “Lora: Low-rank adaptation of large language models,” in ICLR, 2022

    2022

  99. [107]

    Fast inference from transformers via speculative decoding,

    Y. Leviathan, M. Kalman, and Y. Matias, “Fast inference from transformers via speculative decoding,” in ICML, ser. Proceed- ings of Machine Learning Research, vol. 202, 2023, pp. 19 274– 19 286

    2023

  100. [108]

    Unlocking efficiency in large language model inference: A comprehensive survey of speculative decoding,

    H. Xia, Z. Yang, Q. Dong, P . Wang, Y. Li, T. Ge, T. Liu, W. Li, and Z. Sui, “Unlocking efficiency in large language model inference: A comprehensive survey of speculative decoding,” in Findings of ACL, 2024, pp. 7655–7671

    2024

  101. [109]

    Speak from heart: An emotion-guided llm-based multimodal method for emotional dialogue generation,

    C. Liu, Z. Xie, S. Zhao, J. Zhou, T. Xu, M. Li, and E. Chen, “Speak from heart: An emotion-guided llm-based multimodal method for emotional dialogue generation,” in ICMR, 2024, p. 533–542

    2024

  102. [110]

    Cue-CoT: Chain-of-thought prompting for respond- ing to in-depth dialogue questions with LLMs,

    H. Wang, R. Wang, F. Mi, Y. Deng, Z. Wang, B. Liang, R. Xu, and K.-F. Wong, “Cue-CoT: Chain-of-thought prompting for respond- ing to in-depth dialogue questions with LLMs,” in Findings of the Association for Computational Linguistics: EMNLP 2023 , Dec. 2023, pp. 12 047–12 064

    2023

  103. [111]

    Sibyl: Empowering empathetic dialogue generation in large language models via sensible and visionary commonsense inference,

    L. Wang, J. Li, C. Yang, Z. Lin, H. Tang, H. Liu, Y. Cao, J. Wang, and W. Wang, “Sibyl: Empowering empathetic dialogue generation in large language models via sensible and visionary commonsense inference,” in COLING, Jan. 2025, pp. 123–140

    2025

  104. [112]

    TOOL-ED: Enhancing empathetic response generation with the tool calling capability of LLM,

    H. Cao, Y. Zhang, S. Feng, X. Yang, D. Wang, and Y. Zhang, “TOOL-ED: Enhancing empathetic response generation with the tool calling capability of LLM,” in COLING, Jan. 2025, pp. 5305– 5320

    2025

  105. [113]

    Self-consistent reasoning-based aspect-sentiment quad prediction with extract- then-assign strategy,

    J. Kim, R. Heo, Y. Seo, S. Kang, J. Yeo, and D. Lee, “Self-consistent reasoning-based aspect-sentiment quad prediction with extract- then-assign strategy,” in Findings of ACL 2024 , Aug. 2024, pp. 7295–7303. 18 When LLMs Team Up: The Emergence of Collaborative Affective Computing

    2024

  106. [114]

    Distilling fine-grained sentiment understanding from large language models,

    Y. Zhang, G. Xie, H. Xu, K. Hou, J. Bao, Q. Wang, S. Chen, and R. Xu, “Distilling fine-grained sentiment understanding from large language models,” arXiv, 2024

    2024

  107. [115]

    Improving emotional support conver- sation with strategy-intent inference,

    Y. Cao, H. Yu, and F. Wan, “Improving emotional support conver- sation with strategy-intent inference,” in Proceedings of the 2024 8th International Conference on Electronic Information Technology and Computer Engineering, ser. EITCE ’24, 2025, p. 227–233

    2024

  108. [116]

    Exploring chatgpt-based augmentation strategies for contrastive aspect- based sentiment analysis,

    L. Xu, H. Xie, S. J. Qin, F. L. Wang, and X. Tao, “Exploring chatgpt-based augmentation strategies for contrastive aspect- based sentiment analysis,” IEEE Intell. Syst. , vol. 40, no. 1, pp. 69–76, 2025

    2025

  109. [117]

    Iterative data gen- eration with large language models for aspect-based sentiment analysis,

    Q. Zhong, H. Li, L. Zhuang, J. Liu, and B. Du, “Iterative data gen- eration with large language models for aspect-based sentiment analysis,” arXiv, 2024

    2024

  110. [118]

    STICKERCONV: Generating multimodal empathetic responses from scratch,

    Y. Zhang, F. Kong, P . Wang, S. Sun, S. SWangLing, S. Feng, D. Wang, Y. Zhang, and K. Song, “STICKERCONV: Generating multimodal empathetic responses from scratch,” in ACL, Aug. 2024, pp. 7707–7733

    2024

  111. [119]

    SoulChat: Improving LLMs’ empathy, listening, and comfort abilities through fine-tuning with multi-turn empathy conversa- tions,

    Y. Chen, X. Xing, J. Lin, H. Zheng, Z. Wang, Q. Liu, and X. Xu, “SoulChat: Improving LLMs’ empathy, listening, and comfort abilities through fine-tuning with multi-turn empathy conversa- tions,” in Findings of EMNLP, Dec. 2023, pp. 1170–1183

    2023

  112. [120]

    Multi-task learning with llms for implicit sentiment analysis: Data-level and task-level automatic weight learning,

    W. Lai, H. Xie, G. Xu, and Q. Li, “Multi-task learning with llms for implicit sentiment analysis: Data-level and task-level automatic weight learning,” arXiv, 2024

    2024

  113. [121]

    Ckerc : Joint large language models with commonsense knowledge for emotion recognition in conversation,

    Y. Fu, “Ckerc : Joint large language models with commonsense knowledge for emotion recognition in conversation,” arXiv, 2024

    2024

  114. [122]

    Ecr-chain: Advancing generative language models to better emotion-cause reasoners through rea- soning chains,

    Z. Huang, J. Zhao, and Q. Jin, “Ecr-chain: Advancing generative language models to better emotion-cause reasoners through rea- soning chains,” in Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence, IJCAI-24, K. Larson, Ed. Inter- national...

    2024

  115. [123]

    Emollm: Multimodal emotional understanding meets large language models,

    Q. Yang, M. Ye, and B. Du, “Emollm: Multimodal emotional understanding meets large language models,” arXiv, 2024

    2024

  116. [124]

    Llm supervised pre-training for multimodal emotion recognition in conversations,

    S. Dutta and S. Ganapathy, “Llm supervised pre-training for multimodal emotion recognition in conversations,” arXiv, 2025

    2025

  117. [125]

    Affect recognition in conversations using large language models,

    S. Feng, G. Sun, N. Lubis, W. Wu, C. Zhang, and M. Gasic, “Affect recognition in conversations using large language models,” in SIGDIAL, Sep. 2024, pp. 259–273

    2024

  118. [126]

    Wisdom: Improving multimodal sentiment analysis by fusing contextual world knowledge,

    W. Wang, L. Ding, L. Shen, Y. Luo, H. Hu, and D. Tao, “Wisdom: Improving multimodal sentiment analysis by fusing contextual world knowledge,” in ACM MM, 2024, pp. 2282–2291

    2024

  119. [127]

    Designing heterogeneous llm agents for financial senti- ment analysis,

    F. Xing, “Designing heterogeneous llm agents for financial senti- ment analysis,” ACM Trans. Manage. Inf. Syst., vol. 16, no. 1, Feb. 2025

    2025

  120. [128]

    Cooper: Coordinating specialized agents towards a complex dialogue goal,

    Y. Cheng, W. Liu, J. Wang, C. T. Leong, Y. Ouyang, W. Li, X. Wu, and Y. Zheng, “Cooper: Coordinating specialized agents towards a complex dialogue goal,” in AAAI, 2024, pp. 17 853–17 861

    2024

  121. [129]

    Be helpful but don‘t talk too much - enhancing helpfulness in conversations through relevance in multi-turn emotional support,

    J. Li, B. Peng, Y.-Y. Hsu, and C.-R. Huang, “Be helpful but don‘t talk too much - enhancing helpfulness in conversations through relevance in multi-turn emotional support,” in EMNLP, Nov. 2024, pp. 1976–1988

    2024

  122. [130]

    Roberta: A robustly optimized bert pretraining approach,

    Y. Liu, M. Ott, N. Goyal, J. Du, M. Joshi, D. Chen, O. Levy, M. Lewis, L. Zettlemoyer, and V . Stoyanov, “Roberta: A robustly optimized bert pretraining approach,” arXiv, 2019

    2019

  123. [131]

    Sentence-bert: Sentence embed- dings using siamese bert-networks,

    N. Reimers and I. Gurevych, “Sentence-bert: Sentence embed- dings using siamese bert-networks,” in EMNLP-IJCNLP, 2019, pp. 3980–3990

    2019

  124. [132]

    Llama: Open and efficient foundation language models,

    H. Touvron, T. Lavril, G. Izacard, X. Martinet, M. Lachaux, T. Lacroix, B. Rozière, N. Goyal, E. Hambro, F. Azhar, A. Ro- driguez, A. Joulin, E. Grave, and G. Lample, “Llama: Open and efficient foundation language models,” arXiv, 2023

    2023

  125. [133]

    Exploring the limits of transfer learning with a unified text-to-text transformer,

    C. Raffel, N. M. Shazeer, A. Roberts, K. Lee, S. Narang, M. Matena, Y. Zhou, W. Li, and P . J. Liu, “Exploring the limits of transfer learning with a unified text-to-text transformer,” J. Mach. Learn. Res., 2019

    2019

  126. [134]

    Semeval-2014 task 4: Aspect based sentiment analysis,

    M. Pontiki, D. Galanis, J. Pavlopoulos, H. Papageorgiou, I. An- droutsopoulos, and S. Manandhar, “Semeval-2014 task 4: Aspect based sentiment analysis,” in SemEval@COLING, 2014

    2014

  127. [135]

    Llama 2: Open foundation and fine-tuned chat models,

    H. Touvron, L. Martin, K. Stone, P . Albert, A. Almahairi, Y. Babaei, N. Bashlykov, S. Batra, P . Bhargava, S. Bhosale, D. Bikel, L. Blecher, C. C. Ferrer, M. Chen, G. Cucurull, D. Esiobu, J. Fer- nandes, J. Fu, W. Fu, B. Fuller, C. Gao, V . Goswami, N. Goyal, A. Hartshorn, S....

    2023

  128. [136]

    Mixtral of experts,

    A. Q. Jiang, A. Sablayrolles, A. Roux, A. Mensch, B. Savary, C. Bamford, D. S. Chaplot, D. de las Casas, E. B. Hanna, F. Bres- sand, G. Lengyel, G. Bour, G. Lample, L. R. Lavaud, L. Saulnier, M.-A. Lachaux, P . Stock, S. Subramanian, S. Yang, S. Antoniak, T. L. Scao, T. Gervet...

    2024

  129. [137]

    Targeted senti- ment classification with attentional encoder network,

    Y. Song, J. Wang, T. Jiang, Z. Liu, and Y. Rao, “Targeted senti- ment classification with attentional encoder network,” inICANN. Springer-Verlag, 2019, p. 93–103

    2019

  130. [138]

    IEMOCAP: interactive emotional dyadic motion capture database,

    C. Busso, M. Bulut, C. Lee, A. Kazemzadeh, E. Mower, S. Kim, J. N. Chang, S. Lee, and S. S. Narayanan, “IEMOCAP: interactive emotional dyadic motion capture database,” Lang. Resour. Evalu- ation, vol. 42, no. 4, pp. 335–359, 2008

    2008

  131. [139]

    Emotion detection on tv show transcripts with sequence-based convolutional neural networks,

    S. M. Zahiri and J. D. Choi, “Emotion detection on tv show transcripts with sequence-based convolutional neural networks,” ArXiv, vol. abs/1708.04299, 2017

    2017 arXiv

  132. [140]

    MELD: A multimodal multi-party dataset for emo- tion recognition in conversations,

    S. Poria, D. Hazarika, N. Majumder, G. Naik, E. Cambria, and R. Mihalcea, “MELD: A multimodal multi-party dataset for emo- tion recognition in conversations,” in ACL. Association for Computational Linguistics, 2019, pp. 527–536

    2019

  133. [141]

    Chatglm: A family of large language models from glm-130b to glm-4 all tools,

    T. GLM, :, A. Zeng, B. Xu, B. Wang, C. Zhang, D. Yin, D. Zhang, D. Rojas, G. Feng, H. Zhao, H. Lai, H. Yu, H. Wang, J. Sun, J. Zhang, J. Cheng, J. Gui, J. Tang, J. Zhang, J. Sun, J. Li, L. Zhao, L. Wu, L. Zhong, M. Liu, M. Huang, P . Zhang, Q. Zheng, R. Lu, S. Duan, S. Zhang, ...

    2024

  134. [142]

    Bleu: a method for automatic evaluation of machine translation,

    K. Papineni, S. Roukos, T. Ward, and W. Zhu, “Bleu: a method for automatic evaluation of machine translation,” in ACL, 2002, pp. 311–318

    2002

  135. [143]

    ROUGE: A package for automatic evaluation of sum- maries,

    C.-Y. Lin, “ROUGE: A package for automatic evaluation of sum- maries,” in Text Summarization Branches Out, 2004, pp. 74–81

    2004

  136. [144]

    METEOR: an automatic metric for MT evaluation with high levels of correlation with human judg- ments,

    A. Lavie and A. Agarwal, “METEOR: an automatic metric for MT evaluation with high levels of correlation with human judg- ments,” in Proceedings of the Second Workshop on Statistical Machine Translation, WMT@ACL 2007, Prague, Czech Republic, June 23, 2007, 2007, pp. 228–231

    2007

  137. [145]

    A diversity- promoting objective function for neural conversation models,

    J. Li, M. Galley, C. Brockett, J. Gao, and B. Dolan, “A diversity- promoting objective function for neural conversation models,” in NAACL-HLT, 2016, pp. 110–119

    2016

  138. [146]

    Ansari and C

    G. Ansari and C. Saxena, Enhancing Affective Computing in NLP Through Data Augmentation: Strategies for Overcoming Limited Data Availability. Cham: Springer Nature Switzerland, 2024, pp. 201– 216

    2024

  139. [147]

    Learning implicit sentiment in aspect-based sentiment analysis with supervised contrastive pre-training,

    Z. Li, Y. Zou, C. Zhang, Q. Zhang, and Z. Wei, “Learning implicit sentiment in aspect-based sentiment analysis with supervised contrastive pre-training,” in EMNLP, 2021

    2021

  140. [148]

    A survey of confidence estimation and calibration in large language models,

    J. Geng, F. Cai, Y. Wang, H. Koeppl, P . Nakov, and I. Gurevych, “A survey of confidence estimation and calibration in large language models,” in NAACL, Jun. 2024, pp. 6577–6595

    2024

  141. [149]

    Factual confidence of LLMs: on reliability and robustness of current estimators,

    M. Mahaut, L. Aina, P . Czarnowska, M. Hardalov, T. Müller, and L. Marquez, “Factual confidence of LLMs: on reliability and robustness of current estimators,” in ACL, Aug. 2024, pp. 4554– 4570

    2024

  142. [150]

    Recent trends of multimodal affective computing: A survey from nlp perspective,

    G. Hu, Y. Xin, W. Lyu, H. Huang, C. Sun, Z. Zhu, L. Gui, R. Cai, E. Cambria, and H. Seifi, “Recent trends of multimodal affective computing: A survey from nlp perspective,” arXiv, 2024

    2024

  143. [151]

    What makes the difference? an empirical comparison of fusion strategies for multimodal language analysis,

    D. Gkoumas, Q. Li, C. Lioma, Y. Yu, and D. Song, “What makes the difference? an empirical comparison of fusion strategies for multimodal language analysis,” Information Fusion , vol. 66, pp. 184–197, 2021

    2021

  144. [152]

    Are multilingual LLMs culturally-diverse reasoners? an investigation into multi- cultural proverbs and sayings,

    C. Liu, F. Koto, T. Baldwin, and I. Gurevych, “Are multilingual LLMs culturally-diverse reasoners? an investigation into multi- cultural proverbs and sayings,” in NAACL, Jun. 2024, pp. 2016– 2039

    2024

  145. [153]

    Survey of 19 When LLMs Team Up: The Emergence of Collaborative Affective Computing cultural awareness in language models: Text and beyond,

    S. Pawar, J. Park, J. Jin, A. Arora, J. Myung, S. Yadav, F. G. Haznitrama, I. Song, A. Oh, and I. Augenstein, “Survey of 19 When LLMs Team Up: The Emergence of Collaborative Affective Computing cultural awareness in language models: Text and beyond,” arXiv, 2024

    2024

  146. [154]

    Affective computing in education: A systematic review and future research,

    E. Yadegaridehkordi, N. F. B. M. Noor, M. N. B. Ayub, H. B. Affal, and N. B. Hussin, “Affective computing in education: A systematic review and future research,” Comput. Educ., vol. 142, 2019

    2019

  147. [155]

    Ethical considerations on affective computing: An overview,

    L. Devillers and R. Cowie, “Ethical considerations on affective computing: An overview,” Proc. IEEE, vol. 111, no. 10, pp. 1445– 1458, 2023. Wenna Lai (Student Member, IEEE) is currently a Ph.D. student at the Department of Computing, Hong Kong Polytechnic University, under th...

    2023

  148. [2009]

    His research interests include data science, data analytics, recommender systems, web mining, user mod- eling, NLP , social network analysis, and social media mining

    He is currently a Professor and a Program Leader at the School of Computer Science and Data Science Institute, University of Technology Sydney, Sydney, NSW, Australia. His research interests include data science, data analytics, recommender systems, web mining, user mod- eling...

    1982

Pith tools

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