Pith. sign in

REVIEW 5 major objections 6 minor 29 references

IntentionESC: An Intention-Centered Framework for Enhancing Emotional Support in Dialogue Systems

T0 review · 5 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Explicit supporter intention improves emotional support responses in dialogue systems.

desk verdict IntentionESC adds a therapy-grounded intention stage to ESC, but the central causal claim is undercut by derived intention labels and uncontrolled baselines. read the letter →

arxiv 2506.05947 v1 pith:LSPKCI7G submitted 2025-06-06 cs.CL cs.AI

Xinjie Zhang , Wenxuan Wang , Qin Jin This is my paper
classification cs.CLcs.AI
keywords emotionalsupportconversationintentioninferencechain-of-thoughtlargelanguagemodelsstrategyselectionESConvhumanevaluationstateanalysis
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper asks whether emotional support dialogue systems respond better when the supporter's intention is made explicit before a strategy is chosen. It argues that unclear intentions lead supporters to impose their own expectations or solutions, and that a clear intention should drive strategy selection. To test this, the paper defines IntentionESC—a framework with 12 supporter intentions adapted from counseling practice, four aspects of the seeker's emotional state that support intention inference, and a mapping from intentions to strategies—and introduces ICECoT, a chain-of-thought that walks a model through emotional state analysis, intention inference, strategy selection, and response generation. The authors report that a Llama-3.1-8B model trained on auto-annotated ICECoT chains outperforms previous emotional support systems on all of their single-response and whole-conversation evaluation metrics.

What carries the argument

The load-bearing mechanism is the ICECoT reasoning chain together with the intention–strategy mapping from IntentionESC. The mapping connects 12 intentions (e.g., Clarify, Cathart, Insight, Support, Hope, Give information) to the support strategies used in ESConv, such as Open Questions and Probes, Affirmation and Reassurance, and Providing Suggestions. The ICECoT chain makes the reasoning explicit in four steps: emotional state analysis over four aspects, intention inference, strategy selection, and response generation. An automated pipeline built on GPT-4 produces the training chains by refining the original strategy labels and using the intention–strategy mapping to propose intention categories, and the model is trained to reproduce both the chain and the final response.

What would settle it

Run a controlled human evaluation in which the ICECoT model is compared with an identical model whose intention-inference step is replaced by a random intention drawn from the same training distribution, keeping data and generation otherwise unchanged; if scores do not drop significantly, the intention step is not carrying the improvement. A second check is to evaluate ICECoT on emotional support conversations outside ESConv, where the strategy-label distribution that generates the intention annotations is absent.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is that intention is the missing link between understanding the seeker's emotional state and selecting an effective support strategy. The authors define a supporter's intention as the goal, motivation, and expectation behind a response, and they derive 12 such intentions from the therapy literature. They claim that when a model first analyzes the seeker's main issue and causes, current emotions, needs, and relationship dynamics, then infers an intention, then picks a strategy, the resulting responses are more effective than responses generated from emotional state or strategy alone. In their experiments, ICECoT trained on ESConv with automatically annotated reasoning chains receives the best human and GPT-4-based scores on all single-response and entire-conversation dimensions compared to BlenderBot, MultiESC, and ESCoT.

Load-bearing premise

The load-bearing premise is that a supporter's intention can be reliably read off from a seeker's emotional state and, in the training data, from the existing strategy labels; if those strategy labels are noisy or the 12 therapy-derived intentions do not fit everyday support conversations, the intention annotations inherit that bias and the reported gains may not transfer.

Editorial extensions

If this is right

  • Future emotional support systems can condition generation on an inferred intention instead of only on emotion labels or strategies.
  • The four-aspect emotional state checklist provides a reusable annotation scheme for building larger emotional support datasets.
  • ICECoT makes the model's reasoning visible, so users and evaluators can check whether the chosen strategy follows from an intention that matches the seeker's state.
  • Training on mixed data—full reasoning chains plus direct strategy-response pairs—can compensate for fluency losses that long chains cause, a design choice the paper's ablations support.

Reading between the lines

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

  • The 12-intention taxonomy could be validated as a complete ontology by collecting open-domain support dialogues and checking for intentions that fall outside the list; the paper's own limitation about limited scenarios points in this direction.
  • A control condition that replaces the inferred intention with a paraphrase of the selected strategy would reveal whether the intention step adds signal beyond the strategy label, since the annotation pipeline derives intention candidates from strategy annotations.
  • The intention-to-strategy mapping could be exposed to users as a control knob, letting a person choose the kind of support they want (e.g., clarifying questions rather than advice) before the model writes a response.
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

5 major / 6 minor

Summary. The paper proposes IntentionESC, a framework that defines 12 supporter intentions in emotional support conversations, identifies four emotional-state aspects for inferring them, and maps intentions to support strategies. Building on this, the authors introduce ICECoT, a chain-of-thought mechanism in which a model first analyzes the seeker's emotional state, then infers the supporter's intention, then selects a strategy, and finally generates a response. They describe an automated GPT-4-based annotation pipeline that enriches the ESConv dataset with emotional-state and intention labels, train Llama3.1-8B-Instruct on this data, and evaluate the resulting model against BlenderBot, MultiESC, and ESCoT using a mix of GPT-4 ranking and human annotation on single-response and whole-conversation dimensions. The paper claims that ICECoT outperforms all baselines across all evaluated metrics.

Significance. If the central claim were fully supported, the paper would make a useful contribution: explicitly modeling supporter intentions in emotional support dialogue is a plausible route to more interpretable and effective systems, and the released data, code, and evaluation scheme would be reusable assets for the community. The authors also deserve credit for being transparent about the main limitation of their annotation pipeline, for proposing a new evaluation framework that includes safety and consistency dimensions, and for making their data and code available. However, the current evidence is not yet sufficient to establish the claimed superiority of the intention-centered approach: the main baseline comparison is confounded by different base models, the intention labels are derived from the very strategy annotations they are meant to explain, and the reliability of the human evaluation is low on several key dimensions.

major comments (5)
  1. [§5.3, Table 1, Appendix C] The headline comparison in Table 1 is confounded by backbone. ICECoT is trained on Llama3.1-8B-Instruct, while ESCoT is the released Llama2-7B-chat model, and BlenderBot and MultiESC are smaller architectures. Because the main claimed advantage over ESCoT may be explained by the newer/larger instruction-tuned base model, the comparison does not isolate the effect of intention modeling. Please add a same-backbone ESCoT run, or an equivalent Llama3.1 baseline without ICECoT, before claiming that ICECoT outperforms prior methods.
  2. [§4.3, Figure 5, Limitations] The intention labels are not an independent construct. The pipeline first refines ESConv strategy annotations, then generates candidate intentions using the predefined intention–strategy mapping in Figure 2, and then asks GPT-4 to produce a natural-language intention from those candidates. The resulting intention is therefore effectively a paraphrase of a deterministic function of the strategy label that ICECoT is supposed to explain. The Limitations section itself concedes that the pre-existing strategy distribution may bias the intention annotations. To support the central claim, please provide evidence that the learned intention variable carries information beyond the strategy label, for example human intention annotations collected without exposing annotators to the strategy label, or a distributional analysis showing that the emitted intentions are not nearly deterministic given the strategy.
  3. [§5.3, Table 1] The sentence 'Our ICECoT outperforms other compared methods on all metrics' overstates the results. Only four of the seven dimensions carry a ‡ (p<0.05), and the reported p-values appear to come from a single sign test over all models rather than pairwise tests against each baseline. Please report pairwise p-values against each compared method, apply a multiple-comparison correction if needed, and restrict the superiority claim to the dimensions that are actually significant.
  4. [§5.4, Table 2, Appendix E] The full ICECoT chain underperforms direct generation on Base Quality, Empathy, Informativeness, and Comforting. The final best system is ICECoT plus an additional 'strategy-and-response-only' training set D_SA, which was introduced after observing this underperformance. As presented, the central claim that ICECoT itself improves emotional support is therefore contingent on a post-hoc repair. Please justify D_SA as a principled component of the method, report whether its introduction was pre-specified, and provide significance tests comparing the final model with the direct-generation setting on each dimension.
  5. [§5.2, Tables 1 and 2] The inter-annotator agreement is very low for several load-bearing dimensions: in Table 1, Identification κ=0.12 and Comforting κ=0.17; in Table 2, Comforting κ=0.07 and Identification κ=0.21. With agreement at these levels, the ranking data may be too noisy to support fine-grained cross-model conclusions. Please report the number of annotators per item, the adjudication procedure, and consider pooling dimensions or using a more reliable scale, or interpret results with appropriate caution.
minor comments (6)
  1. [§3.1] There is a typo: 'acomplish' should be 'accomplish'.
  2. [Figure 2] The strategy numbers under each intention are difficult to decode; please add a legend that explicitly maps the numbers to the strategy list.
  3. [Tables 1 and 2 captions] The κ values are listed in a single dense line in the captions; please report them per dimension in the table itself so that reliability is transparent and easy to compare.
  4. [§5.2] The composite 'Base Quality' dimension is formed from Fluency, Coherence, Safety, and Consistency, which are evaluated by GPT-4 ranking, while Empathy and Informativeness are evaluated by human annotators; please clarify exactly how the composite ranking was aggregated (e.g., average ranks, majority vote, or another rule).
  5. [Figure 10] The prompt text contains a typo: 'evluation' should be 'evaluation'.
  6. [§1 and §2] The claim of being 'the first to focus on supporters' intentions' should be softened: related work such as GLHG already models a local intention of the current speaker, so the novelty should be positioned as supporter-intention learning and intention-driven strategy selection rather than intention modeling per se.

Circularity Check

1 steps flagged · score 4.0 of 10

Intention labels in §4.3 are generated from ESConv strategy annotations through the paper's own intention–strategy mapping, so the 'intention inference' training signal is a paraphrase of the strategy labels it is claimed to explain; final human evaluation keeps this at partial rather than full circularity.

  1. self definitional [Section 4.3 (Intention Inference), Figure 5; Limitations]
    "(2) Intention Estimation: Using predefined intention-strategy relationships in IntentionESC to estimate possible intentions based on existing strategy annotations; (3) Intention Generation: Producing a final intention annotation for the current response based on dialogue history and candidate intentions."

    The 'intention' labels used to train ICECoT are not obtained from an independent source: they are produced from the ESConv strategy annotations via the paper's own predefined intention–strategy mapping, then paraphrased by GPT-4. So the intention-inference stage is trained to reconstruct a deterministic function of the very strategy labels the framework is supposed to explain. The ICECoT gain over baselines can therefore be explained by earlier injection of strategy-conditioned information, not by an independently meaningful intention construct. The paper's Limitation concedes this ('pre-existing distribution of strategy annotations may disproportionately influence the annotation of intentions').

full rationale

The central empirical comparison (Tables 1–2) evaluates ICECoT against baselines on GPT-4 and human-judged response quality, so the headline result is not a pure tautology and keeps the paper from being fully circular. However, the paper's own annotation pipeline generates intention labels from ESConv strategy annotations using the Figure 2 mapping, making 'intention inference' a paraphrase of strategy labels; the ablation varies whether intention tokens are emitted, but all settings share the same strategy-derived training signal. The Limitations explicitly admit this bias. Because the distinct contribution is the intention construct, and that construct is built out of the very strategy labels it is said to guide, the evidence for the causal role of intention is partially circular. No load-bearing self-citation issue was found: ESCoT is a same-group baseline but functions as a comparison system, not as the justification for the framework's core premise. Score 4 reflects one admitted construct-level reduction with independent final evaluation.

Assumptions & free parameters 0 free parameters · 5 assumptions · 2 invented entities

The framework introduces no numeric fitted parameters, but it rests on hand-specified taxonomies and mappings, on the reliability of GPT-4 annotations, and on the validity of ESConv strategy labels. The intention labels are derived from strategy annotations, which couples training and evaluation to the same prior assumptions.

assumptions (5)
  • domain assumption The 12 intentions adapted from Hill and O'Grady transfer to casual emotional support conversations and map sufficiently well to strategies to be learnable.
    Sections 3.1 to 3.3 adapt therapy intentions to ESC and define the intention-strategy mapping; no empirical validation of the taxonomy itself is reported.
  • domain assumption ESConv strategy annotations are reliable enough to serve as anchors for generating intention labels.
    Section 4.3 pipeline derives intention candidates from existing strategy annotations; the Limitations section explicitly admits this introduces bias risk.
  • domain assumption GPT-4 can produce high-quality emotional state, intention, and strategy annotations following the framework prompts.
    Sections 4.2 to 4.3 and Appendices A to B rely on GPT-4 output; no human validation of annotation quality is reported.
  • domain assumption The lower-rank-mean evaluation with GPT-4 and a few human annotators captures emotional support effectiveness.
    Section 5.2 defines the evaluation; inter-annotator kappas are as low as 0.12 to 0.27, and rankings use small random samples.
  • ad hoc to paper Mixed training data D_SA introduced after observing full ICECoT underperformance is a valid fix rather than an overfit to the evaluation.
    Section 5.4 and Appendix E add strategy-response-only data after observing weaker full-chain results; the change is motivated by observed underperformance, not derived from the framework.
invented entities (2)
  • Supporter intention as an explicit latent variable in emotional support conversations
    purpose: Latent intention labels to guide strategy selection and response generation
    No external ground-truth validation; labels are generated from strategy annotations via GPT-4, and the same framework is used for training and evaluation.
  • ICECoT reasoning chain (emotional state to intention to strategy to response)
    purpose: Structured reasoning path for LLM training and inference
    Internal mechanism; intermediate outputs are evaluated by the framework's own criteria rather than by an external benchmark.

how reviews work

0 comments
Cite this review

Pith. "Pith review of IntentionESC: An Intention-Centered Framework for Enhancing Emotional Support in Dialogue Systems." pith.science (2026). https://pith.science/paper/LSPKCI7G

@misc{pith2026250605947,
  author       = {Pith},
  title        = {Pith review of: IntentionESC: An Intention-Centered Framework for Enhancing Emotional Support in Dialogue Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LSPKCI7G}},
  note         = {Machine review of arXiv:2506.05947}
}
read the original abstract

In emotional support conversations, unclear intentions can lead supporters to employ inappropriate strategies, inadvertently imposing their expectations or solutions on the seeker. Clearly defined intentions are essential for guiding both the supporter's motivations and the overall emotional support process. In this paper, we propose the Intention-centered Emotional Support Conversation (IntentionESC) framework, which defines the possible intentions of supporters in emotional support conversations, identifies key emotional state aspects for inferring these intentions, and maps them to appropriate support strategies. While Large Language Models (LLMs) excel in text generating, they fundamentally operate as probabilistic models trained on extensive datasets, lacking a true understanding of human thought processes and intentions. To address this limitation, we introduce the Intention Centric Chain-of-Thought (ICECoT) mechanism. ICECoT enables LLMs to mimic human reasoning by analyzing emotional states, inferring intentions, and selecting suitable support strategies, thereby generating more effective emotional support responses. To train the model with ICECoT and integrate expert knowledge, we design an automated annotation pipeline that produces high-quality training data. Furthermore, we develop a comprehensive evaluation scheme to assess emotional support efficacy and conduct extensive experiments to validate our framework. Our data and code are available at https://github.com/43zxj/IntentionESC_ICECoT.

Figures

Figures reproduced from arXiv: 2506.05947 by the authors.

Figure 1
Figure 1. Illustration of our IntentionESC framework [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The list of emotional support intentions, definitions, and strategies. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. The keywords in emotional state. It shows the [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figures from the paper (18 more)
Figure 5
Figure 5. Figure 5: The pipeline for automatic intention annotation generation. We use GPT-4 for strategy refinement and intention generation, with the corresponding prompts included in Appendix B. pipeline integrates experiential knowledge of inten￾tion reasoning into actual ESC scenario…
Figure 6
Figure 6. Figure 6: The reasoning process evaluation results. The [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: The prompt for emotional state generation. [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: The prompt for intention generation. We automatically annotate emotional states and emotional support intentions on the ESConv data, incorporating the dataset’s own strategies and re￾sponses to form the ICECoT data. We conduct ex￾periments using LLAMA3.1-8B-Instruct † …
Figure 9
Figure 9. Figure 9: The prompt for strategy refinement. supporter by considering problem identification, comforting skills, and the effectiveness of the sug￾gestions provided, ultimately determining whether a good emotional support experience is delivered. D.2 Base Quality Evaluation To e…
Figure 10
Figure 10. Figure 10: The prompt for basic quality evaluation. [PITH_FULL_IMAGE:figures/full_fig_p014_10.png]
Figure 11
Figure 11. Figure 11: The prompt for profile extraction. You are currently experiencing {situation} and engaging in a conversation for emotional support with a supporter. Below is your personal background: <Personal Summary>: {personal_summary} As a seeker, your task is based on the dialog…
Figure 12
Figure 12. Figure 12: The prompt for seeker simulation. (3) Seeker’s Needs. The score range is [0, 3]. • Is the <Seeker’s Needs> specific and clear? • Does the <Seeker’s Needs> align with the seeker’s expression in the conversation? • Is it possible to identify implicit requirements that t…
Figure 13
Figure 13. Figure 13: The guideline of single response evaluation. [PITH_FULL_IMAGE:figures/full_fig_p016_13.png]
Figure 14
Figure 14. Figure 14: The guideline of entire conversation evaluation. [PITH_FULL_IMAGE:figures/full_fig_p017_14.png]
Figure 15
Figure 15. Figure 15: A case of the emotional support responses by the ablation study. We show the model’s responses in [PITH_FULL_IMAGE:figures/full_fig_p018_15.png]
Figure 16
Figure 16. Figure 16: The second case of our emotional support dialogue system output (ICECoT). [PITH_FULL_IMAGE:figures/full_fig_p019_16.png]
Figure 17
Figure 17. Figure 17: The third case of our emotional support dialogue system output (ICECoT). [PITH_FULL_IMAGE:figures/full_fig_p019_17.png]
Figure 18
Figure 18. Figure 18: The fourth case of our emotional support dialogue system output (ICECoT). [PITH_FULL_IMAGE:figures/full_fig_p020_18.png]
Figure 19
Figure 19. Figure 19: An example of our emotional support dialogue system output (ICECoT), where the output strategy and [PITH_FULL_IMAGE:figures/full_fig_p021_19.png]
Figure 20
Figure 20. Figure 20: The prompt of emotional state analysis evaluation. [PITH_FULL_IMAGE:figures/full_fig_p022_20.png]
Figure 21
Figure 21. Figure 21: The prompt of intention inference evaluation. [PITH_FULL_IMAGE:figures/full_fig_p023_21.png]
Figure 22
Figure 22. Figure 22: The prompt of strategy selection evaluation. [PITH_FULL_IMAGE:figures/full_fig_p023_22.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

29 extracted references · 18 canonical work pages

  1. [1]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRING...

  2. [2]

    write newline

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

  3. [3]

    T Albrecht and M Adelman. 1987. Communicating social support: A theoretical perspective. Communicating Social Support. Beverly Hills, CA: Sage, pages 18--39

    work page 1987

  4. [4]

    Brant R Burleson. 2003. Emotional support skills. In Handbook of communication and social interaction skills, pages 569--612. Routledge

    work page 2003

  5. [5]

    Wei Chen, Hengxu Lin, Qun Zhang, Xiaojin Zhang, Xiang Bai, Xuanjing Huang, and Zhongyu Wei. 2024. Cauesc: A causal aware model for emotional support conversation. arXiv preprint arXiv:2401.17755

    work page Pith review arXiv 2024

  6. [6]

    Yi Chen, Rui Wang, Haiyun Jiang, Shuming Shi, and Ruifeng Xu. 2023. Exploring the use of large language models for reference-free text quality evaluation: An empirical study. In Findings of the Association for Computational Linguistics: IJCNLP-AACL 2023 (Findings), pages 361--374

    work page 2023

  7. [7]

    Jiale Cheng, Sahand Sabour, Hao Sun, Zhuang Chen, and Minlie Huang. 2023. Pal: Persona-augmented emotional support conversation generation. In Findings of the Association for Computational Linguistics: ACL 2023, pages 535--554

    2023

  8. [8]

    Yi Cheng, Wenge Liu, Wenjie Li, Jiashuo Wang, Ruihui Zhao, Bang Liu, Xiaodan Liang, and Yefeng Zheng. 2022. Improving multi-turn emotional support dialogue generation with lookahead strategy planning. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 3014--3026

    2022

Show all 29 references
  1. [9]

    Yi Cheng, Wenge Liu, Jian Wang, Chak Tou Leong, Yi Ouyang, Wenjie Li, Xian Wu, and Yefeng Zheng. 2024. Cooper: Coordinating specialized agents towards a complex dialogue goal. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 17853--17861

    2024

  2. [10]

    Carolyn E Cutrona and Daniel W Russell. 1987. The provisions of social relationships and adaptation to stress. Advances in personal relationships, 1(1):37--67

    1987

  3. [11]

    Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, et al. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783

    2024 arXiv

  4. [12]

    Clara E Hill. 2009. Helping skills: Facilitating, exploration, insight, and action. American Psychological Association

    2009

  5. [13]

    Clara E Hill and Kevin E O'Grady. 2001. List of therapist intentions illustrated in a case study and with therapists of varying theoretical orientations. In Meeting of the Society for Psychotherapy Research., Jun, 1983, Sheffield, England; A version of this study was presented...

    2001

  6. [14]

    Zhaopei Huang, Jinming Zhao, and Qin Jin. 2024. Ecr-chain: Advancing generative language models to better emotion-cause reasoners through reasoning chains. arXiv preprint arXiv:2405.10860

    2024 arXiv

  7. [15]

    Shiyang Li, Jianshu Chen, Zhiyu Chen, Xinlu Zhang, Zekun Li, Hong Wang, Jing Qian, Baolin Peng, Yi Mao, Wenhu Chen, et al. 2024. Explanations from large language models make small reasoners better. In 2nd Workshop on Sustainable AI

    2024

  8. [16]

    Chin-Yew Lin. 2004. Rouge: A package for automatic evaluation of summaries. In Text summarization branches out, pages 74--81

    2004

  9. [17]

    Siyang Liu, Chujie Zheng, Orianna Demasi, Sahand Sabour, Yu Li, Zhou Yu, Yong Jiang, and Minlie Huang. 2021. Towards emotional support dialog systems. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Co...

    2021

  10. [18]

    Yinhong Liu, Han Zhou, Zhijiang Guo, Ehsan Shareghi, Ivan Vuli \'c , Anna Korhonen, and Nigel Collier. 2024. Aligning with human judgement: The role of pairwise preference in large language model evaluators. In First Conference on Language Modeling

    2024

  11. [19]

    Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. Bleu: a method for automatic evaluation of machine translation. In Proceedings of the 40th annual meeting of the Association for Computational Linguistics, pages 311--318

    2002

  12. [20]

    Wei Peng, Yue Hu, Luxi Xing, Yuqiang Xie, Yajing Sun, and Yunpeng Li. 2022. https://doi.org/10.24963/ijcai.2022/600 Control globally, understand locally: A global-to-local hierarchical graph network for emotional support conversation . In Proceedings of the Thirty-First Intern...

    2022 doi

  13. [21]

    Stephen A Rains, Corey A Pavlich, Bethany Lutovsky, Eric Tsetsi, and Anjali Ashtaputre. 2020. Support seeker expectations, support message quality, and supportive interaction processes and outcomes: The case of the comforting computer program revisited. Journal of Social and P...

    2020

  14. [22]

    Stephen Roller, Emily Dinan, Naman Goyal, Da Ju, Mary Williamson, Yinhan Liu, Jing Xu, Myle Ott, Eric Michael Smith, Y-Lan Boureau, and Jason Weston. 2021. https://doi.org/10.18653/v1/2021.eacl-main.24 Recipes for building an open-domain chatbot . In Proceedings of the 16th Co...

    2021 doi

  15. [23]

    Quan Tu, Yanran Li, Jianwei Cui, Bin Wang, Ji-Rong Wen, and Rui Yan. 2022. Misc: A mixed strategy-aware model integrating comet for emotional support conversation. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers...

    2022

  16. [24]

    Lanrui Wang, Jiangnan Li, Chenxu Yang, Zheng Lin, and Weiping Wang. 2023. Enhancing empathetic and emotion support dialogue generation with prophetic commonsense inference. arXiv preprint arXiv:2311.15316

    2023 arXiv

  17. [25]

    Tenggan Zhang, Xinjie Zhang, Jinming Zhao, Li Zhou, and Qin Jin. 2024 a . https://doi.org/10.18653/v1/2024.acl-long.723 ESC o T : Towards interpretable emotional support dialogue systems . In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguisti...

    2024 doi

  18. [26]

    Zaibin Zhang, Yongting Zhang, Lijun Li, Jing Shao, Hongzhi Gao, Yu Qiao, Lijun Wang, Huchuan Lu, and Feng Zhao. 2024 b . Psysafe: A comprehensive framework for psychological-based attack, defense, and evaluation of multi-agent system safety. In Proceedings of the 62nd Annual M...

    2024

  19. [27]

    Weixiang Zhao, Yanyan Zhao, Shilong Wang, and Bing Qin. 2023. Transesc: Smoothing emotional support conversation via turn-level state transition. In Findings of the Association for Computational Linguistics: ACL 2023, pages 6725--6739

    2023

  20. [28]

    Chujie Zheng, Sahand Sabour, Jiaxin Wen, Zheng Zhang, and Minlie Huang. 2023. Augesc: Dialogue augmentation with large language models for emotional support conversation. In Findings of the Association for Computational Linguistics: ACL 2023, pages 1552--1568

    2023

  21. [29]

    Jinfeng Zhou, Zhuang Chen, Bo Wang, and Minlie Huang. 2023. Facilitating multi-turn emotional support conversation with positive emotion elicitation: A reinforcement learning approach. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (...

    2023

Pith tools

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