Emotion-o1: Adaptive Long Reasoning for Emotion Understanding in LLMs

Changhao Song; Hui Gao; Kaiyun Huang; Peng Zhang; Yazhou Zhang

REVIEW 5 major objections 5 minor 1 cited by

Emotion-o1: Adaptive Long Reasoning for Emotion Understanding in LLMs

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 ↗

Pith's one-line read Emotion-o1 claims that letting a model choose how long to reason—short for simple emotions, long for sarcasm and humor—raises F1 by up to 27% and cuts reasoning length by 83%.

desk verdict Useful adaptive-CoT recipe for emotion tasks, but a train/test split ambiguity and metric sloppiness keep the headline gains from being verifiable. read the letter →

arxiv 2505.22548 v2 pith:HUJXVHKK submitted 2025-05-28 cs.CL

Changhao Song , Yazhou Zhang , Hui Gao , Kaiyun Huang , Peng Zhang This is my paper

classification cs.CL

keywords emotionunderstandingchain-of-thoughtreasoningadaptivelengthreinforcementlearningknowledgedistillationsarcasmdetectionhumorsentimentclassification

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper argues that for emotion understanding, a language model's reasoning budget should be per input, not fixed. Emotion-o1 learns to emit short, direct chains of thought for simple tasks like sentiment and longer, reflective chains for pragmatically complex cases like sarcasm and humor. The training pipeline distills variable-length reasoning traces from a strong reasoner, supervises an 8-billion-parameter student on them, then applies reinforcement learning with a reward that balances accuracy, brevity, structure, and redundancy. On four emotion benchmarks the paper reports F1 gains of 10%, 5%, 18%, and 27% over its backbone, while cutting median reasoning length by up to 83% compared with a fixed long-reasoning system. If these results hold, adaptive allocation of reasoning depth is a practical way to get competitive emotion understanding from a small model.

What carries the argument

The load-bearing mechanism is adaptive chain-of-thought length control, where chain-of-thought is the sequence of intermediate reasoning steps a model writes before its final answer. The policy that maps each input to a reasoning trace of appropriate length and structure is induced by distilling a dataset of short/long and linear/non-linear traces from a strong reasoner, keeping only traces whose final label matches the ground truth, fine-tuning on those traces, and then running PPO with a reward that has four terms: accuracy, closeness to quantile-derived length targets, presence of structured reasoning actions (decomposition, reflection, verification), and a penalty for repeated content. The length targets are computed from the distilled trace-length distribution blended with hand-set expectations, so the reward literally encodes the premise that simple emotion tasks should be answered briefly and hard ones should be allowed to think at length.

What would settle it

Take the distilled training set and corrupt the intermediate steps of every trace that ends in the correct label—shuffle the steps or replace them with generic filler—then retrain the same pipeline. If final-label accuracy stays roughly unchanged, the model's gains come from the label-conditioned distribution of the traces rather than from the reasoning content, and the adaptive-reasoning interpretation of the result fails.

Watch

Extended reading notes

Core claim

The central claim is that reasoning depth is a task-specific resource and can be optimized. Emotion-o1 learns to compress reasoning for straightforward cases (sentiment, basic emotion recognition) and to expand it for pragmatically demanding cases (sarcasm, humor). It is built in three stages: rejection-filtered distillation of short/long and linear/non-linear reasoning traces from DeepSeek-R1, supervised fine-tuning on those traces, and PPO reinforcement learning guided by a four-part reward for accuracy, closeness to task-specific length targets, structured reasoning actions, and redundancy. On four benchmarks the paper reports F1 gains of 10%, 5%, 18%, and 27% over its LLaMA-3.1-8B backbone, and reported margins of 1.1% over Grok-3 and 2% over Claude-3.7 on sentiment and sarcasm respectively, while reducing median reasoning length by up to 83% compared with OpenAI-o1. The discovery is that the precision-efficiency trade-off in emotion reasoning can be actively shaped rather than accepted as fixed.

Load-bearing premise

The paper treats a distilled reasoning trace as valid training data whenever it ends at the correct label, so a fluent but internally flawed chain of thought can be baked into the student during fine-tuning and reinforcement learning.

Editorial extensions

If this is right

Fixed-length chain-of-thought is suboptimal for emotion: simple tasks get over-reasoned and complex tasks get under-reasoned, so adaptive length should be the default setting.
An 8-billion-parameter model can reach or beat much larger closed models on sentiment and sarcasm, meaning scale is not the only path to strong emotion understanding.
Reasoning length becomes a trainable efficiency knob, with median reductions of 52-83% relative to a fixed long-reasoning baseline while maintaining accuracy.
The model's learned length distribution reflects task difficulty: sentiment stays short, sarcasm and humor grow long, matching the paper's simple-versus-complex split.

Reading between the lines

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

Beyond the paper, the same adaptive-length recipe should transfer to other classification domains with mixed difficulty, such as hate-speech detection or clinical note triage, because the reward needs only a verifiable label and a length distribution.
A testable extension not run here is a sweep of the expected-length parameter $L_{\text{base}}$ with all other components fixed; if F1 is insensitive to that target, the gains come from the distilled traces, not from length adaptation.
Since the paper itself notes that it considers only text, a natural next test is multimodal emotion input (speech, facial expression): the same reward structure could learn to spend more reasoning on cues that are easy to miss in text.

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.

Desk Editor's Note

read the letter

Quick take: the paper is a plausible adaptive-CoT recipe for emotion tasks, and the gains over the backbone are consistent across four benchmarks. But there's one issue that has to be resolved before I'd trust any number in it: the paper never states whether distillation and the PPO reward use training or test splits. Eq. (2) filters distilled responses by ground-truth label, and Eq. (6) rewards accuracy against ground truth; if test labels enter anywhere in that pipeline, the 10/5/18/27% gains are leakage, not adaptive reasoning. The evaluation section only says "we conduct tasks on the test set," which doesn't tell us what the distillation/RL saw. This is a necessary clarification, not a reproducibility nicety.

What's genuinely new: combining length/structure-distilled CoT with SFT plus a four-part reward for emotion classification is a legit integration. The ablation in Fig. 4 is the most informative part—short CoT matches long on sentiment and emotion, long CoT clearly helps humor and sarcasm. That's a real empirical finding. The efficiency numbers (83% length reduction vs OpenAI-o1) are striking, though they follow partly from the length reward the authors designed; calling this "adaptive" is a bit generous, but the task-dependent gains hold up in the ablation.

Soft spots beyond the split ambiguity: the abstract mixes metrics—the 10% and 5% are Macro-F1/accuracy while the 18% and 27% are Weighted-F1—and that's misleading. Table 1 reports "average results from 3 experiments" but no standard deviations or significance tests. Hyperparameters like alpha, L_exp, lambda, N_A, N_C, and all reward weights are listed as "empirically tuned" but never given. The references include duplicate and extraneous entries, which suggests a hasty compile.

Bottom line: if the split is clean and the metrics are reported consistently, the central claim is likely supportable. As submitted, the paper is a conditional accept at best. It deserves a serious referee round—the idea is useful and the ablation is honest—but the authors need to state splits, report variance, and fix the abstract before I'd believe the headline numbers.

Referee Report

5 major / 5 minor

Summary. The paper presents Emotion-o1, an 8B-parameter LLaMA-3.1-based model for emotion understanding that generates chain-of-thought reasoning whose length adapts to task complexity. Training proceeds in three stages: distillation of variable-length, linear/non-linear reasoning traces from DeepSeek-R1; supervised fine-tuning on those traces; and PPO reinforcement learning with a reward combining accuracy, task-specific length targets, structural keywords, and a repetition penalty. The authors report large F1/accuracy gains over the LLaMA-3.1 backbone on MELD sentiment and emotion recognition, Sarcasm Corpus V2, and Reddit Humor, competitive or superior results against much larger closed-source models, and substantially shorter CoTs than OpenAI-o1 and DeepSeek-R1. The paper claims that Emotion-o1 learns to balance reasoning depth and efficiency automatically.

Significance. If the empirical results hold, the work is significant for practical emotion understanding: an 8B model could approach frontier-level performance with much lower inference cost by allocating reasoning depth per task. The paper also provides a reasonably complete pipeline (distillation, SFT, RL) and a broad comparison against nine baselines, with ablations and length-distribution plots. However, the paper does not release code, data, or hyperparameters, and the central 'adaptive' claim is partly built into the reward design. The significance therefore depends on resolving the data-split and reproducibility concerns below.

major comments (5)

[Structured Emotion Reasoning Distillation / Evaluation Setup] The paper never specifies that the distilled dataset D is constructed from the training splits of the four benchmarks. Since the distillation prompt p(x_i, y_i, c) includes the gold label and Eq. (2) retains only responses whose final label matches y_i, any test instances in D would leak test labels into SFT (Eq. 5) and into the RL accuracy reward (Eq. 6). The only explicit split statement concerns evaluation on the test set. Please state the exact split construction and verify that no test example appears in D or in any reward computation; if this is not already the case, the experiments must be rerun with strict train/test separation.
[Verifiable Reward RL, Eqs. (7)-(9) / CoT Length Analysis] The length reward is task-specific: L_min and L_max are P5/P95 quantiles of the distilled response lengths and L_base is a weighted combination of an author-set L_exp and the median of those same lengths. Consequently, the finding in Table 2 that Sentiment gets shorter CoTs and Sarcasm/Humor get longer ones is substantially determined by the reward, not by the model autonomously discovering task complexity. The authors should provide evidence of adaptivity beyond the encoded targets, such as an ablation without the length reward or an evaluation on unseen emotion tasks, or should soften the 'automatically adjusts' framing.
[Abstract and Main Result] The abstract's 'F1 score increases of 10%, 5%, 18%, and 27%' do not correspond to a single metric in Table 1: the sentiment number is Macro-F1 (+10.1 percentage points), the emotion number is accuracy (+4.9 points) rather than Macro-F1 (+14.5 points) or Weighted-F1 (+10.2 points), and the humor/sarcasm numbers are Macro/Weighted-F1. The paper should report all four tasks with the same metric, and because Table 1 is an average of only 3 runs, standard deviations or confidence intervals are needed to support claims such as '+1.1% over Grok-3' and '+2% over Claude-3.7'.
[Structured Emotion Reasoning Distillation, Eq. (2)] The rejection sampling criterion checks only whether the final label matches the gold label, and the prompt itself provides the gold label, so the distilled traces can be fluent post-hoc rationalizations of a known answer rather than valid reasoning. No human evaluation or consistency check is reported for the intermediate reasoning steps. Without such validation, the improved benchmark scores could reflect better label-conditioned generation rather than improved emotion reasoning, and the interpretability claim is weakened.
[Training Setup / hyperparameters] The method depends on many hyperparameters—alpha, L_exp, L_min/L_max, s_min/s_max/s_base, lambda, N_A/N_C, w_acc/w_length/w_struct/w_repeat, beta, tau, epsilon_acc, and the sampling count N—yet none are reported, and no code or data is provided. This makes the empirical results impossible to reproduce or independently check; please provide the full hyperparameter configuration and release the distillation/filtering scripts and the constructed dataset (or a representative sample) with guaranteed train/test separation.

minor comments (5)

[Introduction and Visualization of CoT Length] The Introduction reports length reductions as '73% (54%), 52% (27%), 83% (70%), 70% (58%)' while the Visualization section gives 83% for Sarcasm and 52% for Emotion; clarify whether these are medians or means, and which baseline (OpenAI-o1 or DeepSeek-R1) each number refers to.
[References] The reference list contains several entries that appear unrelated to the manuscript (e.g., [7]-[10], [13], [17], [22], [28]-[30]) and duplicates ([16] appears to duplicate [11]; [18] duplicates [17]; [29] and [30] duplicate each other); these should be cleaned up.
[Figure 5] The probability density plots are difficult to read; provide the median, percentiles, and standard deviations in tabular form and specify that lengths are measured in tokens.
[Experiments, Training Setup] The statement 'We adopt the LLaMA-3.1-8B as our base model' contains a grammar error; it should read 'We adopt LLaMA-3.1-8B as our base model.'
[Throughout] The model name is inconsistently capitalized; please unify to 'Emotion-o1' everywhere, including the Main Result paragraph where 'emotion-o1' appears.

Circularity Check

2 steps flagged · score 6.0 of 10

Adaptive-length and structure findings largely verify the paper's own reward targets; accuracy gains retain independent content, so the circularity is partial (6/10).

fitted input called prediction [Methodology, 'Verifiable Reward RL' (Eqs. 7-9) and 'CoT Length Analysis' (Table 2)]
"Considering the varying complexity of different emotion tasks, we empirically set an expected lengthL exp based on domain expertise. ... the final expected lengthL base is formulated as a weighted combination of empirical and statistical values:Lbase =α·L exp + (1−α)·L sts. ... These results confirm that our two-stage pipeline enables Emotion-o1 to learn task-aware reasoning strategies, by using shorter CoTs for straightforward tasks and longer ones for cognitively demanding tasks."

The length targets that PPO optimizes are fitted from the authors' own priors and their own distilled data: L_exp is set 'based on domain expertise' (the paper's prior that sentiment/emotion are simple and sarcasm/humor complex), while L_min=P5(L_l), L_max=P95(L_l) and L_sts=median(L_l) are statistics of the paper's own distillation corpus. Eq. (9) then shapes r_length as a Gaussian around L_base with penalties outside [L_min, L_max].
fitted input called prediction [Methodology, 'Verifiable Reward RL' (Eq. 10) and 'Final Reward Function' (Eq. 14)]
"To encourage the model to produce responses that demonstrate a clear and structured reasoning process, we propose a novel structure-oriented rewardr struct. This reward explicitly evaluates key reasoning behaviors (e.g., 'decomposition,' 'reflection,' and 'verification') and the appropriate usage of logical connectives (e.g., 'therefore,' 'however,' and 'thereby')."

Eq. (10) pays the model for containing a hand-coded set of reasoning actions and connectives, and Eq. (14) assigns w_struct its second-highest weight for Long-CoT reasoning. The paper's later conclusions — 'the results validate that sentiment and emotion tasks benefit from concise and direct reasoning, while sarcasm and humor tasks demand longer and deeper reasoning' — restate these reward assignments: behavior that is explicitly rewarded is then reported as evidence that the framework produces structurally coherent, task-appropriate reasoning. The structure findings therefore check that the reward was followed, rather than providing evidence about which reasoning behaviors cause the F1 gains; no RL-stage control without the structure reward is offered.

full rationale

Emotion-o1's central adaptivity claim partially reduces to its own reward construction. The task-aware length reward (Eqs. 7-9) defines per-task expected lengths from (i) author-set 'domain expertise' priors classifying sentiment/emotion as simple and sarcasm/humor as complex, and (ii) quantiles/medians of the paper's own distilled traces. After PPO maximizes this reward, Table 2's shorter Sentiment CoT (85) versus longer Sarcasm (199)/Humor (233) lengths, plus the conclusion that the model 'learned task-aware reasoning strategies,' are a verification of the optimization target by construction — this is the fitted-input-called-prediction pattern. Similarly, the structure reward (Eq. 10) hand-codes preferred behaviors ('decomposition,' 'reflection,' 'verification,' and connectives) that are then reported as the framework's learned structural coherence. The accuracy gains over the backbone (10/5/18/27% F1) are genuine measurements and are not circular, though attributing them to 'adaptive' reasoning is confounded by SFT on DeepSeek-R1-distilled traces that pass a ground-truth-label filter; that is a validity concern, not circularity. Self-citations to the authors' prior work ([39], [41], [42]) exist but are not load-bearing, since the simple/complex task split is also anchored in external references [4, 5]. The most serious unresolved issue is split hygiene: the paper never states that the distillation corpus D (Eq. 3) is drawn from training splits, and the only split mention is 'We conduct tasks on the test set of the benchmark dataset.' If test labels entered Eq. (2) or the Eq. (6) accuracy reward, every reported gain would be forced by construction and the score would be 9-10; because no exhibit permits that reduction, it is flagged as a risk rather than scored as circularity. Overall: partial circularity, 6/10.

Assumptions & free parameters 9 free parameters · 4 assumptions · 0 invented entities

The framework rests on numerous hand-set constants and assumptions about task difficulty and teacher quality. The reward design directly encodes the adaptive behavior that the paper later presents as a finding. No new physical or conceptual entities are introduced.

free parameters (9)

alpha = Not reported
Balance between empirical expected length L_exp and statistical median L_sts in Eq. 8.
L_exp = Not reported
Expected reasoning length per task and reasoning mode, set by domain expertise; directly shapes the length reward.
s_min, s_max, s_base = Not reported
Scaling factors in the length reward (Eq. 9) controlling penalty intensity.
lambda = Not reported
Weight between reasoning actions and logical connectives in the structure reward (Eq. 10).
N_A, N_C = Not reported
Target numbers of reasoning actions and connectives; hand-chosen per task.
w_acc, w_length, w_struct, w_repeat = Not reported
Weights in the final reward (Eq. 14), described as empirically tuned.
beta and tau = Not reported
beta balances lexical and semantic similarity; tau is the repetition threshold in Eqs. 11-13.
epsilon_acc = Not reported
Small penalty for missing predictions in Eq. 6.
N (candidate count) = Not reported
Number of sampled reasoning paths per prompt in Eq. 1, used for rejection sampling.

assumptions (4)

domain assumption Label-matching rejection sampling yields high-quality reasoning traces.
Eq. 2 keeps only responses whose predicted label equals ground truth; intermediate reasoning is not checked for validity.
domain assumption Sentiment and emotion are simple tasks while sarcasm and humor are complex tasks.
This ordering, stated in the Introduction, determines target lengths and reward structure, and is not derived from data.
ad hoc to paper The four reward components accurately capture the goals of emotion reasoning.
Accuracy, length, structure, and repetition rewards are hand-designed; their weights are tuned on the same benchmark tasks, with no external validation.
domain assumption DeepSeek-R1 is a sufficient teacher for emotion reasoning.
All distilled traces come from DeepSeek-R1; no comparison with other teachers or human-written rationales is provided.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Emotion-o1: Adaptive Long Reasoning for Emotion Understanding in LLMs." pith.science (2026). https://pith.science/paper/HUJXVHKK

@misc{pith2026250522548,
  author       = {Pith},
  title        = {Pith review of: Emotion-o1: Adaptive Long Reasoning for Emotion Understanding in LLMs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HUJXVHKK}},
  note         = {Machine review of arXiv:2505.22548}
}

read the original abstract

Long chain-of-thought (CoT) reasoning has shown great promise in enhancing the emotion understanding performance of large language models (LLMs). However, current fixed-length CoT methods struggle to balance reasoning depth and efficiency. Simple tasks (e.g., sentiment classification) are over-reasoned, while complex tasks (e.g., sarcasm understanding) lack depth. To fill this gap, we present Emotion-o1, an adaptive CoT framework that dynamically adjusts reasoning length based on emotion-task complexity. Emotion-o1 is trained by distilling adaptive CoT patterns from a reasoning-oriented LLM, followed by supervised fine-tuning and reinforcement learning with a four-part reward targeting accuracy, brevity, structure, and redundancy. Experimental results on four emotion tasks highlight: (1) Emotion-o1 demonstrates significant improvements over its backbone, with F1 score increases of 10%(Sentiment), 5%(Emotion), 18%(Humor), and 27%(Sarcasm). (2) In sentiment and sarcasm tasks, our 8B model demonstrates superior performance against advanced LLMs, outperforming Grok-3 by 1.1% and Claude-3.7 by 2%. (3) The framework maintains accuracy while reducing reasoning length by 83% compared to OpenAI-o1, demonstrating effective precision-efficiency optimization. Emotion-o1 effectively balances reasoning depth and efficiency for emotion understanding in LLMs.

Figures

Figures reproduced from arXiv: 2505.22548 by the authors.

**Figure 2.** Overview of the proposed framework. Our objective is to jointly model the correctness of l and the interpretability of CoT(x), enabling adaptive reasoning aligned with task difficulty. Structured Emotion Reasoning Distillation We construct annotated samples with diverse reasoning paths by distilling a leading LRM. Specifically, we select four representative emotion understanding tasks, each paired with a widely used… view at source ↗

**Figure 3.** Overview of the CoT Dataset Thus, each textual input xi could yield multiple valid CoT responses across these dimensions, The final dataset D was constructed by aggregating all valid CoT reasoning instances across input samples and reasoning dimensions: D =    (xi, yi, ri,j , ci,j , li,j ) [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Ablation analysis of CoT variants: Long, Short, [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

**Figure 5.** Figure 5: While achieving superior results, Emotion-o1 also [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Towards High-Level Semantic Intelligence
cs.AI 2026-07 conditional novelty 4.0 of 10

A survey proposing that AI's next stage should be understood as High-Level Semantic Intelligence: mastering humor, sarcasm, metaphor, empathy, persuasion, and narrative across modalities.

Reference graph

Works this paper leans on

44 extracted references · 31 canonical work pages · cited by 1 Pith paper

[1]

Will affective computing emerge from foun- dation models and general artificial intelligence? a first evaluation of chatgpt.IEEE Intelligent Systems, 38(2):15–23, 2023

Mostafa M Amin, Erik Cambria, and Bj ¨orn W Schuller. Will affective computing emerge from foun- dation models and general artificial intelligence? a first evaluation of chatgpt.IEEE Intelligent Systems, 38(2):15–23, 2023

work page 2023
[2]

Graph of thoughts: Solving elaborate problems with large language models

Maciej Besta, Nils Blach, Ales Kubicek, Robert Gerstenberger, Michal Podstawski, Lukas Gianinazzi, Joanna Gajda, Tomasz Lehmann, Hubert Niewiadom- ski, Piotr Nyczyk, et al. Graph of thoughts: Solving elaborate problems with large language models. In Proceedings of the AAAI conference on artificial in- telligence, volume 38, pages 17682–17690, 2024

work page 2024
[3]

Language models are few-shot learn- ers.Advances in neural information processing sys- tems, 33:1877–1901, 2020

Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learn- ers.Advances in neural information processing sys- tems, 33:1877–1901, 2020

1901
[4]

Sentiment and emotion help sarcasm? a multi-task learning framework for multi-modal sarcasm, sentiment and emotion analysis

Dushyant Singh Chauhan, SR Dhanush, Asif Ekbal, and Pushpak Bhattacharyya. Sentiment and emotion help sarcasm? a multi-task learning framework for multi-modal sarcasm, sentiment and emotion analysis. InProceedings of the 58th annual meeting of the as- sociation for computational linguistics, pages 4351– 4360, 2020

work page 2020
[5]

A sentiment and emotion aware multimodal multiparty humor recognition in multilingual conversational set- ting

Dushyant Singh Chauhan, Gopendra Vikram Singh, Aseem Arora, Asif Ekbal, and Pushpak Bhattacharyya. A sentiment and emotion aware multimodal multiparty humor recognition in multilingual conversational set- ting. InProceedings of the 29th international confer- ence on computational linguistics, pages 6752–6761, 2022

work page 2022
[6]

Qiguang Chen, Libo Qin, Jiaqi Wang, Jingxuan Zhou, and Wanxiang Che. Unlocking the capabilities of thought: A reasoning boundary framework to quan- tify and optimize chain-of-thought.Advances in Neu- ral Information Processing Systems, 37:54872–54904, 2024

work page 2024
[7]

Clancey.Transfer of Rule-Based Exper- tise through a Tutorial Dialogue

William J. Clancey.Transfer of Rule-Based Exper- tise through a Tutorial Dialogue. Ph.D. diss., Dept. of Computer Science, Stanford Univ., Stanford, Calif., 1979

work page 1979
[8]

William J. Clancey. Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education. InProceedings of the Eighth International Joint Conference on Artificial Intelligence (IJCAI-83), pages 556–560, Menlo Park, Calif, 1983. IJCAI Organization

work page 1983

Show all 44 references

[9]

William J. Clancey. Classification Problem Solving. In Proceedings of the Fourth National Conference on Ar- tificial Intelligence, pages 45–54, Menlo Park, Calif.,
[10]

William J. Clancey. The Engineering of Qualitative Models. Forthcoming, 2021

2021
[11]

DeepSeek-AI, Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qi- hao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, Xiaokang Zhang, Xingkai Yu, Yu Wu, Z. F. Wu, Zhibin Gou, Zhihong Shao, Zhuoshu Li, Ziyi Gao, Aixin Liu, Bing Xue, Bingxuan Wang, Bochao Wu, Bei...

2025
[12]

Bert: Pre-training of deep bidi- rectional transformers for language understanding

Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep bidi- rectional transformers for language understanding. In Proceedings of the 2019 conference of the North Amer- ican chapter of the association for computational lin- guistics: human...

2019
[13]

Blackboard Systems

Robert Engelmore and Anthony Morgan, editors. Blackboard Systems. Addison-Wesley, Reading, Mass., 1986

1986
[14]

Ox- ford University Press, USA, 2002

Dylan Evans.Emotion: The science of sentiment. Ox- ford University Press, USA, 2002

2002
[15]

Cognitive behaviors that enable self-improving reasoners, or, four habits of highly effective stars.arXiv preprint arXiv:2503.01307, 2025

Kanishk Gandhi, Ayush Chakravarthy, Anikait Singh, Nathan Lile, and Noah D Goodman. Cognitive behaviors that enable self-improving reasoners, or, four habits of highly effective stars.arXiv preprint arXiv:2503.01307, 2025

2025 arXiv
[16]

Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learn- ing.arXiv preprint arXiv:2501.12948, 2025

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

2025 arXiv
[17]

Clancey, and Glenn Rennels

Diane Warner Hasling, William J. Clancey, and Glenn Rennels. Strategic explanations for a diagnostic con- sultation system.International Journal of Man- Machine Studies, 20(1):3–19, 1984

1984
[18]

Clancey, Glenn R

Diane Warner Hasling, William J. Clancey, Glenn R. Rennels, and Thomas Test. Strategic Explanations in Consultation—Duplicate.The International Journal of Man-Machine Studies, 20(1):3–19, 1983

1983
[19]

Openai o1 system card.arXiv preprint arXiv:2412.16720, 2024

Aaron Jaech, Adam Kalai, Adam Lerer, Adam Richardson, Ahmed El-Kishky, Aiden Low, Alec Hel- yar, Aleksander Madry, Alex Beutel, Alex Carney, et al. Openai o1 system card.arXiv preprint arXiv:2412.16720, 2024

2024 arXiv
[20]

Rui Mao, Qian Liu, Kai He, Wei Li, and Erik Cambria. The biases of pre-trained language models: An empiri- cal study on prompt-based sentiment analysis and emo- tion detection.IEEE transactions on affective comput- ing, 14(3):1743–1753, 2022

2022
[21]

Gsm-symbolic: Understanding the limitations of math- ematical reasoning in large language models.arXiv preprint arXiv:2410.05229, 2024

Iman Mirzadeh, Keivan Alizadeh, Hooman Shahrokhi, Oncel Tuzel, Samy Bengio, and Mehrdad Farajtabar. Gsm-symbolic: Understanding the limitations of math- ematical reasoning in large language models.arXiv preprint arXiv:2410.05229, 2024

arXiv
[22]

Pluto: The ’other’ red planet

NASA. Pluto: The ’other’ red planet. https://www. nasa.gov/nh/pluto-the-other-red-planet, 2015. Ac- cessed: 2018-12-06

2015
[23]

A review of affective generation models.arXiv preprint arXiv:2202.10763, 2022

Guangtao Nie and Yibing Zhan. A review of affective generation models.arXiv preprint arXiv:2202.10763, 2022

2022 arXiv
[24]

OpenAI, :, Aaron Jaech, Adam Kalai, Adam Lerer, Adam Richardson, Ahmed El-Kishky, Aiden Low, Alec Helyar, Aleksander Madry, Alex Beutel, Alex Carney, Alex Iftimie, Alex Karpenko, Alex Tachard Passos, Alexander Neitz, Alexander Prokofiev, Alexander Wei, Allison Tam, Ally Bennet...

2024
[25]

Creating and characterizing a diverse corpus of sar- casm in dialogue.arXiv preprint arXiv:1709.05404, 2017

Shereen Oraby, Vrindavan Harrison, Lena Reed, Ernesto Hernandez, Ellen Riloff, and Marilyn Walker. Creating and characterizing a diverse corpus of sar- casm in dialogue.arXiv preprint arXiv:1709.05404, 2017

2017 arXiv
[26]

MIT press, 2000

Rosalind W Picard.Affective computing. MIT press, 2000

2000
[27]

Meld: A multimodal multi-party dataset for emotion recognition in conversations.arXiv preprint arXiv:1810.02508, 2018

Soujanya Poria, Devamanyu Hazarika, Navonil Ma- jumder, Gautam Naik, Erik Cambria, and Rada Mi- halcea. Meld: A multimodal multi-party dataset for emotion recognition in conversations.arXiv preprint arXiv:1810.02508, 2018

2018 arXiv
[28]

Poligon: A System for Parallel Problem Solving

James Rice. Poligon: A System for Parallel Problem Solving. Technical Report KSL-86-19, Dept. of Com- puter Science, Stanford Univ., 1986

1986
[29]

Robinson

Arthur L. Robinson. New ways to make microcircuits smaller.Science, 208(4447):1019–1022, 1980

1980
[30]

Robinson

Arthur L. Robinson. New Ways to Make Microcircuits Smaller—Duplicate Entry.Science, 208:1019–1026, 1980

1980
[31]

Instruction tuning for large language models: A survey.arXiv preprint arXiv:2308.10792, 2023

Zhang Shengyu, Dong Linfeng, Li Xiaoya, Zhang Sen, Sun Xiaofei, Wang Shuhe, Li Jiwei, Runyi Hu, Zhang Tianwei, Fei Wu, et al. Instruction tuning for large language models: A survey.arXiv preprint arXiv:2308.10792, 2023

2023
[32]

A survey of sentiment analysis: Approaches, datasets, and future research.Applied Sciences, 13(7):4550, 2023

Kian Long Tan, Chin Poo Lee, and Kian Ming Lim. A survey of sentiment analysis: Approaches, datasets, and future research.Applied Sciences, 13(7):4550, 2023

2023
[33]

Qwq-32b: Embracing the power of rein- forcement learning, March 2025

Qwen Team. Qwq-32b: Embracing the power of rein- forcement learning, March 2025

2025
[34]

Gomez, Lukasz Kaiser, and Illia Polosukhin

Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. Attention is all you need, 2017

2017
[35]

Tech- niques of sarcasm detection: A review

Palak Verma, Neha Shukla, and AP Shukla. Tech- niques of sarcasm detection: A review. In2021 inter- national conference on advance computing and inno- vative technologies in engineering (ICACITE), pages 968–972. IEEE, 2021

2021
[36]

Chain-of-thought prompting elicits reasoning in large language models.Advances in neural informa- tion processing systems, 35:24824–24837, 2022

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

2022
[37]

Humor detection: A transformer gets the last laugh.”Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing”, November 2019

Orion Weller and Kevin Seppi. Humor detection: A transformer gets the last laugh.”Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing”, November 2019

2019
[38]

Tokenskip: Controllable chain-of- thought compression in llms, 2025

Heming Xia, Chak Tou Leong, Wenjie Wang, Yongqi Li, and Wenjie Li. Tokenskip: Controllable chain-of- thought compression in llms, 2025

2025
[39]

Is sarcasm detection a step-by-step reasoning process in large language models? InProceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 25651–25659, 2025

Ben Yao, Yazhou Zhang, Qiuchi Li, and Jing Qin. Is sarcasm detection a step-by-step reasoning process in large language models? InProceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 25651–25659, 2025

2025
[40]

Tree of thoughts: Deliberate prob- lem solving with large language models, 2023.URL https://arxiv

Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Thomas L Griffiths, Yuan Cao, and Karthik Narasimhan. Tree of thoughts: Deliberate prob- lem solving with large language models, 2023.URL https://arxiv. org/abs/2305.10601, 3:1, 2023

2023 arXiv
[41]

Cmma: benchmarking multi- affection detection in chinese multi-modal conversa- tions.Advances in Neural Information Processing Sys- tems, 36:18794–18805, 2023

Yazhou Zhang, Yang Yu, Qing Guo, Benyou Wang, Dongming Zhao, Sagar Uprety, Dawei Song, Qi- uchi Li, and Jing Qin. Cmma: benchmarking multi- affection detection in chinese multi-modal conversa- tions.Advances in Neural Information Processing Sys- tems, 36:18794–18805, 2023

2023
[42]

Sarcasmbench: Towards evaluat- ing large language models on sarcasm understanding, 2024

Yazhou Zhang, Chunwang Zou, Zheng Lian, Prayag Tiwari, and Jing Qin. Sarcasmbench: Towards evaluat- ing large language models on sarcasm understanding, 2024

2024
[43]

Both matter: Enhancing the emotional in- telligence of large language models without com- promising the general intelligence.arXiv preprint arXiv:2402.10073, 2024

Weixiang Zhao, Zhuojun Li, Shilong Wang, Yang Wang, Yulin Hu, Yanyan Zhao, Chen Wei, and Bing Qin. Both matter: Enhancing the emotional in- telligence of large language models without com- promising the general intelligence.arXiv preprint arXiv:2402.10073, 2024

2024 arXiv
[44]

Large language models are human-level prompt engineers

Yongchao Zhou, Andrei Ioan Muresanu, Ziwen Han, Keiran Paster, Silviu Pitis, Harris Chan, and Jimmy Ba. Large language models are human-level prompt engineers. InThe eleventh international conference on learning representations, 2022

2022

Pith tools

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

[1] [1]

Will affective computing emerge from foun- dation models and general artificial intelligence? a first evaluation of chatgpt.IEEE Intelligent Systems, 38(2):15–23, 2023

Mostafa M Amin, Erik Cambria, and Bj ¨orn W Schuller. Will affective computing emerge from foun- dation models and general artificial intelligence? a first evaluation of chatgpt.IEEE Intelligent Systems, 38(2):15–23, 2023

work page 2023

[2] [2]

Graph of thoughts: Solving elaborate problems with large language models

Maciej Besta, Nils Blach, Ales Kubicek, Robert Gerstenberger, Michal Podstawski, Lukas Gianinazzi, Joanna Gajda, Tomasz Lehmann, Hubert Niewiadom- ski, Piotr Nyczyk, et al. Graph of thoughts: Solving elaborate problems with large language models. In Proceedings of the AAAI conference on artificial in- telligence, volume 38, pages 17682–17690, 2024

work page 2024

[3] [3]

Language models are few-shot learn- ers.Advances in neural information processing sys- tems, 33:1877–1901, 2020

Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learn- ers.Advances in neural information processing sys- tems, 33:1877–1901, 2020

1901

[4] [4]

Sentiment and emotion help sarcasm? a multi-task learning framework for multi-modal sarcasm, sentiment and emotion analysis

Dushyant Singh Chauhan, SR Dhanush, Asif Ekbal, and Pushpak Bhattacharyya. Sentiment and emotion help sarcasm? a multi-task learning framework for multi-modal sarcasm, sentiment and emotion analysis. InProceedings of the 58th annual meeting of the as- sociation for computational linguistics, pages 4351– 4360, 2020

work page 2020

[5] [5]

A sentiment and emotion aware multimodal multiparty humor recognition in multilingual conversational set- ting

Dushyant Singh Chauhan, Gopendra Vikram Singh, Aseem Arora, Asif Ekbal, and Pushpak Bhattacharyya. A sentiment and emotion aware multimodal multiparty humor recognition in multilingual conversational set- ting. InProceedings of the 29th international confer- ence on computational linguistics, pages 6752–6761, 2022

work page 2022

[6] [6]

Qiguang Chen, Libo Qin, Jiaqi Wang, Jingxuan Zhou, and Wanxiang Che. Unlocking the capabilities of thought: A reasoning boundary framework to quan- tify and optimize chain-of-thought.Advances in Neu- ral Information Processing Systems, 37:54872–54904, 2024

work page 2024

[7] [7]

Clancey.Transfer of Rule-Based Exper- tise through a Tutorial Dialogue

William J. Clancey.Transfer of Rule-Based Exper- tise through a Tutorial Dialogue. Ph.D. diss., Dept. of Computer Science, Stanford Univ., Stanford, Calif., 1979

work page 1979

[8] [8]

William J. Clancey. Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education. InProceedings of the Eighth International Joint Conference on Artificial Intelligence (IJCAI-83), pages 556–560, Menlo Park, Calif, 1983. IJCAI Organization

work page 1983