From Generic Empathy to Personalized Emotional Support: A Self-Evolution Framework for User Preference Alignment

Chengqing Zong; Jing Ye; Lu Xiang; Yaping Zhang

REVIEW 5 major objections 5 minor 26 references

From Generic Empathy to Personalized Emotional Support: A Self-Evolution Framework for User Preference Alignment

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

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

Pith's one-line read This paper shows that a 7B chatbot learns to give more personalized and diverse emotional support by treating its own self-refined response as the preferred answer over its original draft, via iterative direct preference optimization.

desk verdict A competent, applied self-evolution-for-personalized-emotional-support paper whose central personalization claim is provisional because the preference labels are self-referential and the human evaluation is too thin. read the letter →

arxiv 2505.16610 v1 pith:FF2MCA3I submitted 2025-05-22 cs.CL

Jing Ye , Lu Xiang , Yaping Zhang , Chengqing Zong This is my paper · ORCID

classification cs.CL

keywords emotionalsupportconversationself-evolutiondirectpreferenceoptimizationself-refinementuseralignmentpersonalizedempathyLLM

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 claims that large language models give generic, formulaic empathy because they never consider who the user is, what they are feeling, or what situation they are in. It proposes a two-phase self-evolution framework: first, fine-tune on limited emotional support conversations to get basic skills; second, have the model reflect on the user's profile, emotion, and situation, refine its own draft response, and train with direct preference optimization, treating the refined response as chosen and the original as rejected. Across LLaMA-3-8B, Qwen2-7B, and Mistral-7B, the twice-iterated model (M2) outperforms the fine-tuned baseline (M0) on automatic metrics and, in pairwise human evaluation, wins 54.1% of comparisons against M0. This matters because it suggests that alignment with implicit user preferences can come from the model's own reflection on a conversation, without extra labels or reward models, and without explicit reflection at inference time.

What carries the argument

The engine is a self-refinement preference pair combined with iterative direct preference optimization (DPO). Given dialogue history Cn, the model drafts a response y_tn; then, prompted with human-designed principles I, it infers the user's profile un and emotional state sn, and produces a refined response \tilde{y}_t^n. The pair (y_tn, \tilde{y}_t^n) becomes the rejected/chosen pair in the DPO loss (Eq. 6), with an added SFT term on the chosen response (Eq. 7) using γ = 1. Iterating the loop—using Mt−1 to generate the next iteration's preference data—is what makes improvement continue beyond M1. A data-filtering pipeline substitutes golden responses when refined responses exceed twice the length of the draft or fail JSON parsing, which controls length bias but also reveals that the chosen label is not always reliable on its own.

What would settle it

A held-out human annotation study of the preference pairs themselves: if human judges, blind to which response is which, do not prefer the self-refined responses over the originals more often than chance (on the same dialogue contexts), the preference signal is invalid, and the reported gains would be artifacts of style or length rather than better alignment.

Watch

Extended reading notes

Core claim

The central claim is that the gap between generic empathy and personalized emotional support can be closed by using the model's own self-refinement as a preference signal. In the first phase, the model is fine-tuned (via LoRA) on the manually annotated ESConv dataset plus instruction-following data to acquire basic emotional support skills. In the second phase, from a dialogue context, the current model Mt first drafts a response, then summarizes the user's profile un and emotional state sn, and then produces a refined response guided by those reflections. The pair (draft, refined) is used as (rejected, chosen) in the DPO objective, with a small SFT term on chosen responses to stabilize training. Iterating this yields M1 and M2; the paper reports that M2 shows clear improvements over M0 in coherence, distinct-n diversity, and human-rated engagement, informativeness, and helpfulness across all three 7B backbones. The authors also show that fixed human or GPT-4-generated chosen responses give diminishing gains across iterations, while self-refined chosen responses keep improving, making the self-refined preference data the load-bearing ingredient of continued improvement.

Load-bearing premise

The load-bearing premise is that a model's self-refined response is genuinely better for the user than its original draft, since the chosen/rejected labels in Eq. (6) are written by the same model, guided only by a prompt listing strategies, with no external reward or human verification per sample.

Editorial extensions

If this is right

Three 7B backbones (LLaMA-3-8B, Qwen2-7B, Mistral-7B) all improve from M0 to M2, indicating the mechanism is not tied to one architecture.
Because the trained model no longer needs the reflection prompt at inference, the aligned model is cheaper to deploy than online self-refinement methods that require multi-turn inference.
Iterative training on self-generated preference data keeps improving from M1 to M2, whereas fixed human or GPT-4 chosen responses plateau, so the framework offers a path for continued self-evolution as data are regenerated.
The self-evolution training raises distinct-n diversity and BLEU scores over M0, reducing formulaic, repetitive empathy while preserving general reasoning and instruction-following abilities under LoRA.
The 54.1% pairwise human win rate of M2 against M0 suggests that users notice the difference in personalization, not just a change in wording.

Reading between the lines

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

An implicit corollary is that the same self-refinement-as-preference trick may transfer to other subjective generation tasks—tutoring, negotiation, counseling—where a user's implicit preference is unlabeled, provided the self-refinement prompt encodes the right dimensions.
A further inference is that the method's ceiling is bounded by the base model's reflection ability: models too weak to refine their own drafts would produce a label whose noise DPO can only amplify, so gains should shrink as backbone capability decreases.
The length-cap and golden-response substitutions in the filtering pipeline imply that the purity of the chosen/rejected labels is already doubtful; a controlled study measuring label noise (for example, what fraction of refined responses are shorter or are near-paraphrases of the original) would tell whether DPO is learning alignment or only a stylistic shift.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Quick take: this is a competent application of the self-refinement-plus-DPO recipe to emotional support, and it shows real promise. The new piece is the domain-specific reflection on user profile, situation, and emotion, used to create chosen responses for iterative DPO. The authors demonstrate consistent—if modest—objective gains across three 7B backbones, and their small interactive human evaluation prefers M2 over M0. It deserves a careful referee, but not before the evaluation gaps are addressed.

What I found genuinely useful is the ablation in Table 5. Comparing human-written, GPT-4-generated, and self-refined chosen responses within the same DPO setup is a clean test of the framework's central bet, and the result—self-refined data keeps improving across iterations while the fixed human/GPT-4 data plateaus—is the best evidence in the paper that self-refinement is doing something beyond copying a static teacher. The general-capability checks in Table 4 are also a good habit, and the citation pattern is honest: they build directly on the self-evolution and DPO work they cite.

The load-bearing weakness is that the preference signal is self-referential. The same model writes the rejected draft, writes the chosen refinement from a hand-written prompt, and is then trained to prefer its own refinement. Nothing external verifies that 'refined' means 'better for the user.' The Appendix A filters control length, JSON errors, and greeting turns; they do not validate semantic preference. The stress-test concern about style shift is fair—in Table 5's SR row, BLEU-2 rises while Distinct-2 falls 13.29% from M1 to M2, which is exactly the pattern you'd expect from optimizing a narrow stylistic signal. That doesn't mean it happened; the 54.1% pairwise human win is external evidence pointing in the right direction. But four graduate students with no significance test is a thin reed for the headline claim, and the LLM-as-judge evaluation likely shares the same stylistic preferences.

The tables need a cleanup pass. Table 5's M0 row appears to be the vanilla backbone, not the SFT-based M0 used everywhere else, and some Table 6 percentages don't match the numbers they derive from. To the authors' credit, they explicitly acknowledge the evaluation limitations in the Limitations section.

Who this is for: researchers working on ESC personalization or self-evolving dialogue models. The recipe is simple and reproducible in principle, and the component choices are well grounded. I'd treat the personalization claim as provisional until code, seeds, and a larger human evaluation with error bars are available.

Recommendation: yes, send it to peer review; a serious referee can sort out evaluation standards. But ask for released code and cleaned statistics before it becomes a strong citation.

Referee Report

5 major / 5 minor

Summary. This paper proposes a two-phase self-evolution framework for personalized emotional support. In the first phase, a backbone LLM is fine-tuned on ESConv with LoRA to obtain M0. In the second phase, M0 generates unrestricted responses, then a hand-written reflection prompt guides it to summarize the user's profile, emotion, and situation and refine the response; the original and refined responses form rejected/chosen pairs for DPO, with an SFT term on the chosen responses, iterated to M1 and M2. The paper reports automatic metrics on the ESConv test set across LLaMA-3-8B, Qwen2-7B, and Mistral-7B, plus interactive pointwise and pairwise human evaluations and a GPT-4o-as-a-judge evaluation, claiming that M2 yields more diverse, personalized, and user-aligned responses while removing the need for explicit reflection at inference.

Significance. The core idea—using a model's own self-refined output as the preferred response and its original output as the dispreferred response in iterative DPO—is simple, cost-effective, and potentially useful for aligning LLMs with user preferences without external labels. Strengths include the multi-backbone comparisons, the interactive human evaluation protocol, and the transparency about data filtering and limitations. However, the central claim that this procedure learns genuine user preferences is not yet established because the preference labels are self-generated and never externally validated, and the evaluation lacks the statistical rigor needed to support the strength of the claims. If these issues are addressed with additional validation and analysis, the work could be a valuable contribution to LLM alignment and emotional support dialogue.

major comments (5)

[Section 2.2.1, Eq. (5)-(6)] The DPO training treats every self-refined response as chosen and the model's original draft as rejected, but the refined response is produced by the same model Mt guided only by the hand-written prompt in Appendix E.1, with no external reward or human verification per sample. The Appendix A filters enforce length, parsing, and greeting constraints, not semantic preference validity. This is the load-bearing assumption of the method. The Limitations section acknowledges 'potential biases and noise,' but the main text provides no evidence that refined responses are actually preferable for users. I would like to see a human-validated sample of preference pairs (e.g., 100 pairs per iteration) with agreement rates, or a demonstration that an external reward model agrees with the self-refinement direction. Without this, the reported gains could reflect the model amplifying stylistic artifacts of the reflection prompt rather than true alignment with user preferences.
[Section 4.1, Tables 2 and 6] The paper claims 'significantly improves' and 'clear improvements' across backbones, but no standard deviations, confidence intervals, or significance tests are reported for any automatic metric. The differences are often small (e.g., BLEU-2 18.38 to 20.06 for LLaMA) and could be within run-to-run variation; Table 6 shows that M1-to-M2 changes are sometimes negative (e.g., LLaMA BLEU-2 -0.79%). To support the cross-backbone generalization claim, the authors should report results over multiple seeds or provide bootstrap confidence intervals, and ideally a significance test (e.g., paired bootstrap or t-test) for the key comparisons.
[Section 5.3, Table 5] The row labeled 'M0' reports BLEU-2=11.29, ROUGE-L=10.43, and Distinct-2=72.83, which are exactly the Vanilla model values from Table 2, not the M0 values (18.38, 16.72, 90.84) that correspond to the SFT-based M0 defined in Section 3.3. This mislabeling confounds the ablation study: the '{SR,P}' condition is evaluated from the untuned backbone rather than from the M0 model used in the main experiments, so the claim that self-refinement 'consistently improves across successive refinement iterations' is not supported by the reported base condition. Please correct the label or rerun the ablation from the correct M0.
[Section 4.1 and Section 5.3, Tables 2 and 5] The metric trends are internally inconsistent with the claimed improvements in coherence and consistency. In Table 2, ROUGE-L declines from M0 to M2 for all three backbones (e.g., LLaMA: 16.72 to 15.50). In Table 5, the {SR,P} pair shows Distinct-2 dropping by 13.29% from M1 to M2 while BLEU-2 rises by 9.66%. This pattern is consistent with the model learning a narrow stylistic shift (e.g., shorter responses, strategy-word patterns) rather than broad quality improvement. The paper should discuss these conflicting metrics directly or provide additional analyses (e.g., length-controlled generation, per-strategy breakdown, or a qualitative study of where the gains come from) to rule out style artifacts.
[Section 4.2, Figure 4] The pairwise human evaluation uses only four graduate students, and no significance test or per-rater breakdown is reported. The paper gives only percentage win/tie/lose values without stating the number of pairwise judgments. With only four raters, the win rate for M2 vs M0 (54.1%) may not be distinguishable from chance, and the results are sensitive to individual rater preferences. Please report the exact number of comparisons, per-rater win rates, inter-rater agreement, and a significance test. The pointwise evaluation with 50 students is more substantial, but the paper states that inter-rater reliability was not computed; for a central claim of user preference alignment, this is a major gap.

minor comments (5)

[Section 3.4.1] Typo: 'Appenix F' should be 'Appendix F'.
[Section 2.2.1, Eqs. (4)-(5)] Notation inconsistency: Eq. (4) defines (un, sn), but the surrounding text and Eq. (5) refer to the user profile as pn; please make the notation consistent.
[Section 5.2, Figure 5b] The caption says 'increasing correlation between chosen and rejected responses,' but this is actually a similarity measure; please specify the exact metric and what it measures.
[Appendix A] The rule that removes the first turn and the last two turns assumes they are greetings without presenting supporting evidence; this ad-hoc filter could remove meaningful emotional content and should be justified with data or an analysis.
[Section 5.3, Table 5 caption] The column abbreviations 'B-2', 'B-3', 'R-l', 'D-2' are not defined in the caption; define them (BLEU-2, BLEU-3, ROUGE-L, Distinct-2) for readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the DPO training labels are self-generated, but the paper's evaluation is external to those labels.

full rationale

The only self-referential element is the construction of preference pairs: Eq. (3) generates the rejected response y_t^n = M_t(C_n), and Eq. (5) generates the chosen response as the same model's self-refined output \tilde{y}_t^n = M_t(I || C_n, u_n, s_n, y_t^n); Eq. (6) then trains M_{t+1} to prefer the refined output. This is a self-training loop, not a circular derivation: the chosen label is an assumption about quality, and the paper does not define 'personalized support' as 'output of Eq. (5)'. The claim is tested against independent evidence—held-out ESConv reference responses (Table 2), interactive human pointwise and pairwise evaluations (Figures 3-4), and general LLM benchmarks (Table 4). The 54.1% pairwise win of M2 over M0 is an external judgment. The paper's Limitations section explicitly concedes that the synthetic preference data 'introduces potential biases and noise,' which is an acknowledged correctness risk rather than a hidden circularity. The only self-citation of note is ServeForEmo (Ye et al., 2024) used as one of two seed datasets; it is a data resource, not a load-bearing theorem, and the results do not reduce to it. No equation or fitted parameter is renamed as a prediction, and no uniqueness claim is imported from prior author work. Hence no significant circularity.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The framework contributes a training loop, not a theory. It imports DPO, LoRA, and self-refinement assumptions, and adds hand-chosen thresholds and filtering rules. The self-refinement-as-ground-truth assumption is the largest unpaid debt because each sample's chosen response is not independently verified. No new physical or conceptual entities are introduced.

free parameters (4)

DPO temperature beta = 0.1
Set by hand in the final loss L = L_DPO + gamma * L_SFT; controls the strength of preference optimization and is load-bearing for training stability.
SFT loss weight gamma = 1.0
Set by hand alongside beta; balances the DPO term and the SFT term in Eq. (8).
Length ratio threshold for chosen-response substitution = 2x rejected or golden response length
In Appendix A, refined responses longer than twice the rejected or golden response are replaced by the golden response to control length bias; this arbitrary threshold shapes the preference data.
Greeting-turn removal rule = first turn and last two turns removed
Appendix A removes the first turn and last two turns of each dialogue based on prior knowledge, altering the training distribution without an ablation.

assumptions (5)

standard math Direct preference optimization on a dataset of chosen and rejected responses moves the model toward the distribution of chosen responses.
The framework inherits DPO's assumptions from Rafailov et al. 2023; Eq. (6) is used without modification.
ad hoc to paper An LLM's self-refined response, produced with a hand-written reflection prompt, is genuinely preferable for the user than its original draft.
This is the core generative assumption in Section 2.2.1 and Eq. (5); there is no per-sample external verification, and if false the preference labels are invalid.
domain assumption The LLM can reliably infer user profile, emotional state, and implicit preferences from dialogue context.
Invoked in Eq. (4) and in the self-refinement prompt; justified by citations to prior LLM capabilities rather than measured in this paper.
domain assumption Replay with 500 Alpaca instruction samples preserves general abilities when combined with LoRA fine-tuning.
Section 2.1.2 and Table 4 support the preservation claim empirically, but the causal role of replay is not isolated.
ad hoc to paper Removing greeting turns and applying length caps improves preference data quality without introducing bias.
Appendix A describes these filtering rules, but no ablation isolates their effect on the final model.

how reviews work

0 comments

Cite this review

Pith. "Pith review of From Generic Empathy to Personalized Emotional Support: A Self-Evolution Framework for User Preference Alignment." pith.science (2026). https://pith.science/paper/FF2MCA3I

@misc{pith2026250516610,
  author       = {Pith},
  title        = {Pith review of: From Generic Empathy to Personalized Emotional Support: A Self-Evolution Framework for User Preference Alignment},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FF2MCA3I}},
  note         = {Machine review of arXiv:2505.16610}
}

read the original abstract

Effective emotional support hinges on understanding users' emotions and needs to provide meaningful comfort during multi-turn interactions. Large Language Models (LLMs) show great potential for expressing empathy; however, they often deliver generic and one-size-fits-all responses that fail to address users' specific needs. To tackle this issue, we propose a self-evolution framework designed to help LLMs improve their responses to better align with users' implicit preferences concerning user profiles (personalities), emotional states, and specific situations. Our framework consists of two distinct phases: \textit{(1)} \textit{Emotional Support Experience Acquisition}, where LLMs are fine-tuned on limited emotional support conversation data to provide basic support, and \textit{(2)} \textit{Self-Improvement for Personalized Emotional Support}, where LLMs leverage self-reflection and self-refinement to generate personalized responses. Through iterative direct preference optimization between the pre- and post-refined responses, our model generates responses that reflect a better understanding of the user's implicit preferences. Extensive experiments and evaluations demonstrate that our method significantly enhances the model's performance in emotional support, reducing unhelpful responses and minimizing discrepancies between user preferences and model outputs.

Figures

Figures reproduced from arXiv: 2505.16610 by the authors.

**Figure 1.** Example responses. Direct prompting of LLaMA results in verbose and formulaic outputs. Task-Specific SFT is empathetic but often lacks depth and variety, giving it a perceived "AI-like" quality. In contrast, self-reflection on user preferences provides a pathway to more specific and engaging responses. contextually appropriate emotional support is essential for ensuring practical assistance (Campos et al., 2018; Ch… view at source ↗

**Figure 2.** The overview of our self-evolution framework, which enhances personalized emotional support capabilities through a [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Interactive pointwise human evaluation results. The [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Figures from the paper (9 more)

**Figure 5.** Figure 5: (a) Distribution of response relevance to user state [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

**Figure 6.** Figure 6: Comparison of the frequent phrases in responses [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]

**Figure 7.** Figure 7: Example case dialogue generated by M0 -LLaMA (Relationship Problem) [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]

**Figure 8.** Figure 8: Example case dialogue generated by M2 -LLaMA (Relationship Problem) [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]

**Figure 9.** Figure 9: Example case dialogue generated by M0 -LLaMA (Academic Problem) [PITH_FULL_IMAGE:figures/full_fig_p018_9.png]

**Figure 10.** Figure 10: Example case dialogue generated by M2 -LLaMA (Academic Problem) [PITH_FULL_IMAGE:figures/full_fig_p019_10.png]

**Figure 11.** Figure 11: The interface of the interactive point-wise human evaluation. [PITH_FULL_IMAGE:figures/full_fig_p026_11.png]

**Figure 12.** Figure 12: The interface of the interactive pair-wise human evaluation. [PITH_FULL_IMAGE:figures/full_fig_p026_12.png]

**Figure 13.** Figure 13: The detailed guidelines for human evaluation. [PITH_FULL_IMAGE:figures/full_fig_p027_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

26 extracted references · 21 canonical work pages

[3]

In Proceedings of the 2024 Conference on Empirical Methods in Natu- ral Language Processing, pages 1976–1988, Miami, Florida, USA

Be helpful but don’t talk too much - enhanc- ing helpfulness in conversations through relevance in multi-turn emotional support. In Proceedings of the 2024 Conference on Empirical Methods in Natu- ral Language Processing, pages 1976–1988, Miami, Florida, USA. Association for Computational Lin- guistics. R Likert. 1932. A technique for the measurement of a...

arXiv 2024
[7]

It sounds like

OpenReview.net. Zhiyuan Zeng, Jiatong Yu, Tianyu Gao, Yu Meng, Tanya Goyal, and Danqi Chen. 2024. Evaluating large lan- guage models at evaluating instruction following. In The Twelfth International Conference on Learning Representations. Tenggan Zhang, Xinjie Zhang, Jinming Zhao, Li Zhou, and Qin Jin. 2024. Escot: Towards interpretable emotional support ...

work page 2024
[8]

Prompt for w/ self-reflection You are an emotional support expert

Others: Exchanging pleasantries or offering other emotional support. Prompt for w/ self-reflection You are an emotional support expert. You can use the following strategies to engage with users:

work page
[16]

Question: Asking for information related to the problem to help the seeker articulate the issues that they face

work page
[23]

Before responding to the user, please follow these steps:

Others: Exchanging pleasantries or offering other emotional support. Before responding to the user, please follow these steps:

work page
[25]

Select a Strategy: Choose a response strategy based on the user’s emotional needs and prefer- ences

work page
[26]

Respond: Respond to the user with an appropriate message based on the selected strategy. Your answer should be formatted as a JSON block: { 'strategy': <one of the strategies>, 'text': <your response> } Prompt for generating chosen response You are an emotional support expert. You can use the following strategies to engage with users: 1. Question: Asking ...

work page
[27]

Affirmation and Reassurance: Offering reassurance and affirming the help-seeker’s feelings or experiences

work page

Show all 26 references

[28]

Reflection of Feelings: Articulating and describing the seeker’s feelings
[29]

Information: Providing useful information, such as data, facts, opinions, or resources, or answering questions
[30]

Providing Suggestions: Offering suggestions on how to approach the issue, without overstepping or telling them what to do
[31]

Restatement or Paraphrasing: Rephrasing the help-seeker’s statements more concisely to help them see the situation clearly
[32]

Self-disclosure: Sharing similar experiences or emotions to express empathy with the help- seeker
[33]

Your task is to evaluate the target sys’s response and refine it

Others: Exchanging pleasantries or offering other emotional support. Your task is to evaluate the target sys’s response and refine it. For each target sys’s response:
[34]

Understand the User: Understand the user’s profile, characteristics, emotional needs, and potential preferences they reveal in the conversation
[35]

Evaluate the Response: Rate the target system response on a scale of 1-5 based on how well it meets the user’s needs, aligns with their preferences, and provides appropriate emotional support
[36]

Provide Feedback: Identify specific weaknesses in the original response, such as tone, empathy level, or relevance, and explain how it could be improved to better support the user
[37]

Refine the Response: Provide a revised version of target system’s response that better aligns with the user’s emotional needs and preferences. Your answer should be formatted as a JSON block inside markdown: { 'understanding': { 'user_profile': <the profile of the user>, 'user...
[38]

Strategy application alignment with the conversation stage
[39]

Emotional perceptiveness
[40]

Don’t feel that way

Response quality control Apply strict penalties for critical errors. [Evaluation Guideline] A. Strategy Implementation Evaluate the proper use of 8 emotional support strategies:[Question, Affirmation and Reassurance, Reflection of Feelings, Information, Providing Suggestions, ...
[585]

Jiashuo Wang, Chunpu Xu, Chak Tou Leong, Wenjie Li, and Jing Li

Association for Computational Linguistics. Jiashuo Wang, Chunpu Xu, Chak Tou Leong, Wenjie Li, and Jing Li. 2024b. Muffin: Mitigating unhelp- fulness in emotional support conversations with mul- tifaceted AI feedback. In Findings of the Association for Computational Linguistic...

2024 arXiv
[2018]

In Proceedings of the 17th International Conference on Autonomous Agents and MultiAgent Systems, pages 1649–1657

Challenges in exploiting conversational mem- ory in human-agent interaction. In Proceedings of the 17th International Conference on Autonomous Agents and MultiAgent Systems, pages 1649–1657. Yi Chen, Rui Wang, Haiyun Jiang, Shuming Shi, and Ruifeng Xu. 2023. Exploring the use ...

2023 arXiv
[2022]

Advances in neural in- formation processing systems, 35:27730–27744

Training language models to follow instruc- tions with human feedback. Advances in neural in- formation processing systems, 35:27730–27744. Kishore Papineni, Salim Roukos, Todd Ward, and Wei- Jing Zhu. 2002. Bleu: a method for automatic evalu- ation of machine translation. In ...

2002 arXiv
[2023]

Direct preference optimization: Your language model is secretly a reward model. In Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Sys- tems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023. Stephen A Rai...

2023 arXiv
[2024]

Sarik Ghazarian, Ralph M

A framework for few-shot language model evaluation. Sarik Ghazarian, Ralph M. Weischedel, Aram Galstyan, and Nanyun Peng. 2019. Predictive engagement: An efficient metric for automatic evaluation of open- domain dialogue systems. CoRR, abs/1911.01456. Geyang Guo, Ranchi Zhao, ...

2019 arXiv

Pith tools

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

[1] [3]

In Proceedings of the 2024 Conference on Empirical Methods in Natu- ral Language Processing, pages 1976–1988, Miami, Florida, USA

Be helpful but don’t talk too much - enhanc- ing helpfulness in conversations through relevance in multi-turn emotional support. In Proceedings of the 2024 Conference on Empirical Methods in Natu- ral Language Processing, pages 1976–1988, Miami, Florida, USA. Association for Computational Lin- guistics. R Likert. 1932. A technique for the measurement of a...

arXiv 2024

[2] [7]

It sounds like

OpenReview.net. Zhiyuan Zeng, Jiatong Yu, Tianyu Gao, Yu Meng, Tanya Goyal, and Danqi Chen. 2024. Evaluating large lan- guage models at evaluating instruction following. In The Twelfth International Conference on Learning Representations. Tenggan Zhang, Xinjie Zhang, Jinming Zhao, Li Zhou, and Qin Jin. 2024. Escot: Towards interpretable emotional support ...

work page 2024

[3] [8]

Prompt for w/ self-reflection You are an emotional support expert

Others: Exchanging pleasantries or offering other emotional support. Prompt for w/ self-reflection You are an emotional support expert. You can use the following strategies to engage with users:

work page

[4] [16]

Question: Asking for information related to the problem to help the seeker articulate the issues that they face

work page

[5] [23]

Before responding to the user, please follow these steps:

Others: Exchanging pleasantries or offering other emotional support. Before responding to the user, please follow these steps:

work page

[6] [25]

Select a Strategy: Choose a response strategy based on the user’s emotional needs and prefer- ences

work page

[7] [26]

Respond: Respond to the user with an appropriate message based on the selected strategy. Your answer should be formatted as a JSON block: { 'strategy': <one of the strategies>, 'text': <your response> } Prompt for generating chosen response You are an emotional support expert. You can use the following strategies to engage with users: 1. Question: Asking ...

work page

[8] [27]

Affirmation and Reassurance: Offering reassurance and affirming the help-seeker’s feelings or experiences

work page