Evaluating the Performance of Large Language Models in Scientific Claim Detection and Classification

Tanjim Bin Faruk

REVIEW 4 major objections 4 minor 21 references

Evaluating the Performance of Large Language Models in Scientific Claim Detection and Classification

T0 review · 4 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read GPT-4, prompted but not fine-tuned, detects and classifies COVID-19 scientific claims in tweets better than GPT-3.5 and three Llama 2 models, the paper reports.

desk verdict The paper's central claim collapses on its own tables: GPT-3.5's Verifiable F1 is 0.87 in the text and 0.70 in Table 5, and the correct number determines whether GPT-4 actually wins. read the letter →

arxiv 2412.16486 v1 pith:WM4G4ARN submitted 2024-12-21 cs.CL cs.CYcs.LGcs.SI

Tanjim Bin Faruk This is my paper · ORCID

classification cs.CLcs.CYcs.LGcs.SI

keywords LargeLanguageModelsscientificclaimdetectionCOVID-19misinformationfact-checkingGPT-4Llama2Twitterdatasetpromptengineering

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 sets out to show that off-the-shelf large language models can spot scientific claims about COVID-19 in tweets and judge whether those claims are verifiable, all without task-specific training. Using 1,847 manually annotated tweets and prompt-engineering techniques like few-shot examples and chain-of-thought reasoning, the study reports GPT-4 as the top performer, with $F_1$ scores of 0.65 for claim existence and 0.76 for verifiability. Because the setup requires no fine-tuning, the authors argue LLMs could be deployed quickly as automated fact-checking tools during public-health crises.

What carries the argument

The load-bearing mechanism is the two-part label scheme of claim existence and scientific verifiability, taken from an existing annotation framework, combined with system prompts that direct the models to act as a 'COVID-19 Tweets Classifier.' The prompts mount few-shot examples, explicit guidelines, emotional cues, chain-of-thought steps, and clue-reasoning under a single deterministic decoding configuration (temperature 0.2, top-p 0.4), extracting labels directly from generative LLMs without any fine-tuning.

What would settle it

Run each model multiple times on the same 1,847 tweets with different random seeds at the same temperature, compute 95% confidence intervals for the $F_1$ scores; if the intervals for GPT-4 and GPT-3.5 overlap, the paper's claim that GPT-4 'outshone' the others collapses to 'statistically indistinguishable.'

Watch

Extended reading notes

Core claim

In the paper's own terms, the central discovery is that GPT-4 outperforms every other tested model on both tasks, reaching an $F_1$ of 0.65 for deciding whether a tweet contains a scientific claim and 0.76 for deciding whether that claim is verifiable against scientific evidence. GPT-3.5 is close behind (0.61 and 0.70), while the Llama 2 family trails, with the 70B model surprisingly weaker on verifiability (0.63) than the 7B and 13B variants (0.72 and 0.71). Across all models, precision runs higher than recall, so the systems rarely raise false alarms but routinely miss actual claims.

Load-bearing premise

The measured differences between models are treated as real even though each model was run once with no repeated runs, no confidence intervals, and no significance tests, so the 0.04–0.06 $F_1$ gaps between GPT-4, GPT-3.5, and the top Llama could be sampling or API randomness.

Editorial extensions

If this is right

A prompt-only LLM pipeline could categorize COVID-19 tweets in hours, far faster than training a supervised classifier.
The measured ranking suggests model size is not a reliable predictor: the 70B Llama scored lowest on verifiability.
Deployments would benefit from recall-boosting post-processing because all models miss many true claims.
The successful use of an established annotation framework means existing claim datasets can be reused as few-shot examples for new LLM applications.

Reading between the lines

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

A single evaluation pass with fixed decoding settings cannot distinguish genuine model differences from API nondeterminism; rerunning GPT-3.5 and GPT-4 with multiple seeds would test whether the reported $F_1$ gaps are stable.
Extending the best model with retrieval-augmented generation, as the paper's future-work section suggests, could convert a claim detector into a claim verifier that cites the scientific literature for each flagged claim.
For public-health practice, the low recall numbers imply these tools are best used as a first-pass triage to queue tweets for human review, rather than as an autonomous filter.

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

Read this as a course-project report, not a research paper. The headline result—GPT-4 outshines all models on COVID-19 claim detection—is undercut by a direct internal contradiction. Section 7 says GPT-3.5's Verifiable F1 is 0.87; Table 5 reports 0.70. Recomputing from the precision and recall in the table gives 0.71, so the table value is internally consistent and the text's 0.87 is likely an error. But this isn't a minor typo: if 0.87 is right, GPT-3.5 beats GPT-4's 0.76 and the abstract's conclusion is false; if 0.70 is right, GPT-3.5 ranks below the Llama-2 7B and 13B models, and the text's 'outperformed the Llama 2 models' is false. The paper never acknowledges or resolves this.

What the paper does well: it uses an externally annotated dataset (Zuo et al., 2022) with the SciTweets annotation framework, so there's no circularity. It documents five prompt variants and the decoding hyperparameters (temperature 0.2, top-p 0.4, etc.) in enough detail to reproduce. It also reports precision, recall, and F1, which is better than only accuracy. The observation that Llama 2 7B and 13B beat the 70B on the Verifiable task is counterintuitive and genuinely interesting if it survives a proper protocol.

The soft spots are load-bearing. The evaluation is a single run: no repeated trials, no error bars, no significance tests. The gaps between GPT-4 and GPT-3.5 are 0.04–0.06 F1, which is exactly the kind of difference that sampling or API nondeterminism can produce. The paper never says how scores from the five prompt versions were combined—one prompt? best-of? average?—so the reported numbers are ambiguous even setting the 0.87/0.70 issue aside. No code or raw predictions are released, and Table 2 has a typo ('0.0.72'). These are fixable, but they mean the paper's only real finding—the model ranking—is not currently supported by the evidence.

Who this is for: someone wanting a compact example of prompt templates for claim classification, or a pointer to the dataset. Not for anyone needing reliable evidence about LLM capability.

My recommendation: don't publish as is. The experiment is cheap to repair: rerun with multiple seeds, report mean and standard deviation, specify prompt aggregation, correct the internal inconsistency, and release the predictions. A fixed version might be acceptable for a workshop or as a negative result in a broader study. As submitted, it doesn't meet the standard for a serious peer-reviewed venue, but the underlying work isn't worthless—it needs to be redone carefully.

Referee Report

4 major / 4 minor

Summary. The paper evaluates five large language models (GPT-3.5, GPT-4, Llama 2 7B/13B/70B) on two binary classification tasks derived from a published, manually annotated COVID-19 tweet dataset: detecting whether a claim exists (Claim Existence) and whether it is scientifically verifiable (Verifiable). Five system-prompt variants are described, and per-model accuracy, precision, recall, and F1 are tabulated. The central claim is that GPT-4 achieves the highest F1 on both tasks and that LLMs have significant potential as automated fact-checking tools.

Significance. If the reported comparison were reliable, the paper would offer a modest empirical data point on zero-shot claim detection and classification by current LLMs on a specialized, manually annotated dataset. The manuscript is transparent about its dataset source, model families, and decoding hyperparameters, which are strengths. However, the reliability of the headline ranking is undermined by an internal numerical contradiction in the GPT-3.5 Verifiable F1 value, the unspecified protocol for aggregating results across five prompt variants, and the complete absence of variance or significance analysis. These issues make the core comparative claim unsupported as stated.

major comments (4)

[Section 7 and Table 5] The text of the GPT 3.5 paragraph states 'a significantly better performance in the Verifiable task with an F1-score of 0.87', but Table 5 reports F1 = 0.70 for the same task. Recomputing F1 from Table 5's precision (0.77) and recall (0.66) gives 2*0.77*0.66/(0.77+0.66) ≈ 0.711, which supports the table and contradicts the text. This inconsistency is load-bearing: if 0.70 is correct, the sentence 'It outperformed the Llama 2 models in detecting verifiable claims' is false (Llama 2 7B and 13B both have higher Verifiable F1); if 0.87 is correct, then GPT-4's Verifiable F1 of 0.76 is not the highest, so the claim that 'GPT 4 outshone all other models' is false. The paper does not resolve this contradiction, and no raw predictions or code are provided, so the reader cannot determine which value is intended.
[Section 5.1 and Section 7] The paper introduces five distinct system prompt versions (FSP, Guidelines+FSP, Guidelines+FSP+Emotional Stimuli, Chain of Thought, CARP) but reports only a single set of metrics per model-task in Section 7. It is never specified whether the reported numbers correspond to the best-performing prompt per model, an average across prompts, or some other aggregation rule. Because prompt design is known to have a large effect on LLM classification performance, the absence of an aggregation protocol makes the results impossible to reproduce and prevents the reader from assessing whether the reported ranking is robust to prompt choice.
[Sections 7 and 8] All reported metrics derive from a single evaluation pass with no repeated runs, confidence intervals, or statistical significance tests. In this setup, the differences between GPT-4 and GPT-3.5 are small: 0.04 F1 on Claim Existence (0.65 vs 0.61) and, under Table 5, 0.06 F1 on Verifiable (0.76 vs 0.70). Given the nondeterminism inherent in LLM decoding and API serving (even at temperature 0.2), these gaps are within plausible random fluctuation. The paper's assertion that GPT-4 'outshone' other models is therefore not statistically justified.
[Entire paper (Related Work and Results)] The paper motivates LLMs as superior to traditional machine learning models, citing SciBERT- and RoBERTa-based detectors in the related work, but it includes no non-LLM baseline in the experiments. Without comparing against such a baseline, the paper cannot support its broader claim that LLMs 'bypass' overfitting or offer a more adaptable solution. The empirical contribution is limited to a comparison among a small set of LLMs, and the stronger positioning in the abstract and introduction is untested.

minor comments (4)

[Table 2] The F1 entry for the Verifiable task is written as '0.0.72'; this appears to be a typographical error that should read '0.72'.
[Section 5.1] The prompt code blocks contain LaTeX macro artifacts such as 'Var' and 'quotedbl.Var', which render the example prompts as corrupted and difficult to read. These should be cleaned up for presentation.
[Section 5.3] The paper lists both Top P = 0.4 and Top K = 50 in the parameter settings but does not state whether both were applied jointly or whether one overrode the other. Many inference APIs treat top-p and top-k as alternative sampling strategies, so the actual decoding behavior is ambiguous.
[Section 4.3] The model descriptions lack precise version identifiers (e.g., which GPT-3.5 snapshot or Llama 2 checkpoint was used), which is important for reproducibility since provider-side updates can change behavior.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the evaluation is an external benchmark against published manual annotations, with no fitted parameters or self-citation chain.

full rationale

The paper is an empirical benchmark study. Its central measurements are LLM predictions compared against human-annotated labels from a previously published dataset [18]; the ground truth is not produced by the paper's own models, equations, or fitted parameters. The prompt engineering reuses the annotation framework of [3] as an input condition, but the model still independently classifies held-out tweets, so the reported F1 scores are not equal to the prompt definitions by construction. No parameter is fitted to the evaluation data, no prediction is a renamed fit, and no load-bearing claim is justified solely by a self-citation. The internal inconsistency between the Section 7 text (GPT-3.5 Verifiable F1=0.87) and Table 5 (F1=0.70) is a serious correctness and reproducibility problem, but it is not a circularity problem: it does not show that an output is equivalent to an input. Accordingly, no circular step can be identified under the stated criteria.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The evaluation depends on the reliability of external manual labels and on the stability of model outputs under the chosen prompt and decoding scheme. No new entities are introduced; the only hand-chosen numbers are the decoding hyperparameters, which are fixed across models but unvaried in sensitivity analysis.

free parameters (1)

Decoding hyperparameters = temperature=0.2, top_p=0.4, top_k=50, repetition_penalty=1, max_new_tokens=1500
Chosen by hand and fixed across all models; no sensitivity analysis is provided, so the ranking is conditional on these settings.

assumptions (3)

domain assumption Manual annotations in the dataset are treated as reliable ground truth for claim existence and verifiability.
Section 4.2.2 states two annotators plus arbitration but reports no inter-annotator agreement; the SciTweets criteria include subjective categories such as absurdity and humor, so label reliability is assumed.
domain assumption The evaluation protocol, including a single run and fixed decoding parameters, yields stable and comparable metrics across models.
Section 5.3 fixes decoding parameters, but no repeats or variance estimates are reported; Section 7 reports only point estimates.
ad hoc to paper The five prompt versions jointly represent each model's capability, rather than the best prompt being selected per model.
Section 5.1 lists five prompts but never states whether final scores are averages, maxima, or per-prompt; this selection choice can bias the comparison.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Evaluating the Performance of Large Language Models in Scientific Claim Detection and Classification." pith.science (2026). https://pith.science/paper/WM4G4ARN

@misc{pith2026241216486,
  author       = {Pith},
  title        = {Pith review of: Evaluating the Performance of Large Language Models in Scientific Claim Detection and Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WM4G4ARN}},
  note         = {Machine review of arXiv:2412.16486}
}

read the original abstract

The pervasive influence of social media during the COVID-19 pandemic has been a double-edged sword, enhancing communication while simultaneously propagating misinformation. This \textit{Digital Infodemic} has highlighted the urgent need for automated tools capable of discerning and disseminating factual content. This study evaluates the efficacy of Large Language Models (LLMs) as innovative solutions for mitigating misinformation on platforms like Twitter. LLMs, such as OpenAI's GPT and Meta's LLaMA, offer a pre-trained, adaptable approach that bypasses the extensive training and overfitting issues associated with traditional machine learning models. We assess the performance of LLMs in detecting and classifying COVID-19-related scientific claims, thus facilitating informed decision-making. Our findings indicate that LLMs have significant potential as automated fact-checking tools, though research in this domain is nascent and further exploration is required. We present a comparative analysis of LLMs' performance using a specialized dataset and propose a framework for their application in public health communication.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

21 extracted references · 4 canonical work pages

[1]

Leticia Bode and Emily K. Vraga. 2018. See Something, Say Something: Correction of Global Health Misinforma- tion on Social Media. Health Communication 33, 9 (2018), 1131–1140. h/t_tps://doi.org/10.1080/10410236.2017.1331312 arXiv:https://doi.org/10.1080/10410236.2017.1331312 PMID: 28622038

arXiv 2018
[2]

Curran, Allyson Gallant, Helen Wong, Ca ther- ine Johnson, Alannah Delahunty-Pike, Lynora Saxinger, Derek Chu, Jeannette Comeau, Trudy Flynn, Julie Clegg, and Christopher Dye

Leah Boulos, Janet A. Curran, Allyson Gallant, Helen Wong, Ca ther- ine Johnson, Alannah Delahunty-Pike, Lynora Saxinger, Derek Chu, Jeannette Comeau, Trudy Flynn, Julie Clegg, and Christopher Dye

work page
[3]

Salim Haﬁd, Sebastian Schellhammer, Sandra Bringay, Konstantin Todorov, and Stefan Dietze. 2022. SciTweets - A Dataset and An- notation Framework for Detecting Scientiﬁc Online Discourse. In Proceedings of the 31st ACM International Conference on Inf ormation & Knowledge Management (Atlanta, GA, USA) (CIKM ’22). Associ- ation for Computing Machinery, New ...

arXiv 2022
[4]

Beizhe Hu, Qiang Sheng, Juan Cao, Yuhui Shi, Yang Li, Dand- ing Wang, and Peng Qi. 2023. Bad Actor, Good Advisor: Explor- ing the Role of Large Language Models in Fake News Detection. arXiv: 2309.12247 [cs.CL]

arXiv 2023
[5]

Yue Huang and Lichao Sun. 2023. Harnessing the Power of ChatGPT in Fake News: An In-Depth Exploration in Generation, Detection and Explanation. arXiv: 2310.05046 [cs.CL]

arXiv 2023
[6]

Rohit Kumar Kaliyar, Anurag Goswami, and Pratik Narang. 2021. FakeBERT: Fake news detection in social media with a BERT-based deep learning approach. Multimedia Tools and Applications 80, 8 (Jan. 2021), 11765–11788. h/t_tps://doi.org/10.1007/s11042-020-10183-2

doi:10.1007/s11042-020-10183-2 2021
[7]

Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, Sebastian Riedel, and Douwe Kiela
[8]

Cheng Li, Jindong Wang, Yixuan Zhang, Kaijie Zhu, Wenxin Hou, Jianxun Lian, Fang Luo, Qiang Yang, and Xing Xie. 2023. Large Lan- guage Models Understand and Can be Enhanced by Emotional Stimuli. arXiv: 2307.11760 [cs.CL]

arXiv 2023

Show all 21 references

[9]

Mancenido, and Huan Liu

Ahmadreza Mosallanezhad, Mansooreh Karami, Kai Shu, Michel le V. Mancenido, and Huan Liu. 2022. Domain Adaptive Fake News De- tection via Reinforcement Learning. In Proceedings of the ACM Web Conference 2022 (Virtual Event, Lyon, France) (WWW ’22). Associ- ation for Computing ...

2022
[10]

Navendu Prakash, Bhavya Srivastava, Shveta Singh, Seema Sh arma, and Sonali Jain. 2022. Eﬀectiveness of social distancing interven- tions in containing COVID-19 incidence: International evidence us- ing Kalman ﬁlter. Economics & Human Biology 44 (2022), 101091. h/t_tps://doi.o...

2022
[11]

Arkadiy Saakyan, Tuhin Chakrabarty, and Smaranda Muresan. 20 21. COVID-Fact: Fact Extraction and Veriﬁcation of Real-World Claim s on COVID-19 Pandemic. arXiv: 2106.03794 [cs.CL]

arXiv
[12]

Giovanni Spitale, Nikola Biller-Andorno, and Federico Germani. 202 3. AI model GPT-3 (dis)informs us better than humans. Science Ad- vances 9, 26 (2023), eadh1850. h/t_tps://doi.org/10.1126/sciadv.adh1850 arXiv:https://www.science.org/doi/pdf/10.1126/sciadv.adh1850

2023 doi
[13]

Xiaofei Sun, Xiaoya Li, Jiwei Li, Fei Wu, Shangwei Guo, Tianwei Zhang, and Guoyin Wang. 2023. Text Classiﬁcation via Large Lan- guage Models. arXiv: 2305.08377 [cs.CL]

2023 arXiv
[14]

David Wadden, Kyle Lo, Bailey Kuehl, Arman Cohan, Iz Beltagy , Lucy Lu Wang, and Hannaneh Hajishirzi. 2022. SciFact-Open: To- wards open-domain scientiﬁc claim veriﬁcation. In Findings of the As- sociation for Computational Linguistics: EMNLP 2022 . Association for Computation...

2022
[15]

Amelie Wührl and Roman Klinger. 2021. Claim Detection in Biomed- ical Twitter Posts. CoRR abs/2104.11639 (2021). arXiv: 2104.11639 h/t_tps://arxiv.org/abs/2104.11639

2021 arXiv
[16]

Xueyao Zhang, Juan Cao, Xirong Li, Qiang Sheng, Lei Zhong, and Kai Shu. 2021. Mining Dual Emotion for Fake News Detection. In Proceed- ings of the Web Conference 2021 (Ljubljana, Slovenia) (WWW ’21). As- sociation for Computing Machinery, New York, NY, USA, 3465–3476. h/t_tps:...

2021
[17]

Yongchun Zhu, Qiang Sheng, Juan Cao, Shuokai Li, Danding Wang, and Fuzhen Zhuang. 2022. Generalizing to the Fu- ture: Mitigating Entity Bias in Fake News Detection. In Pro- ceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retri...

2022
[18]

acm-jdslogo.png

Chaoyuan Zuo, Ritwik Banerjee, Fateme Hashemi Chaleshtor i, Hos- sein Shirazi, and Indrakshi Ray. 2022. Seeing Should Probably Not Be Believing: The Role of Deceptive Support in COVID-19 Misinforma- tion on Twitter. J. Data and Information Quality 15, 1, Article 9 (dec 2022), ...

2022 arXiv
[2020]

In Proceedings of the 34th International Conference on Neural In- formation Processing Systems (Vancouver, BC, Canada) (NIPS’20)

Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. In Proceedings of the 34th International Conference on Neural In- formation Processing Systems (Vancouver, BC, Canada) (NIPS’20). Cur- ran Associates Inc., Red Hook, NY, USA, Article 793, 16 pages
[2023]

Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineerin g Sciences 381, 2257 (2023), 20230133

Eﬀectiveness of face masks for reducing transmission of SARS-CoV-2: a rapid systematic review. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineerin g Sciences 381, 2257 (2023), 20230133. h/t_tps://doi.org/10.1098/rsta.2023.0133 arXiv:https:/...

2023
[4734]

h/t_tps://doi.org/10.18653/v1/2022.findings-emnlp.347

2022 doi

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [1]

Leticia Bode and Emily K. Vraga. 2018. See Something, Say Something: Correction of Global Health Misinforma- tion on Social Media. Health Communication 33, 9 (2018), 1131–1140. h/t_tps://doi.org/10.1080/10410236.2017.1331312 arXiv:https://doi.org/10.1080/10410236.2017.1331312 PMID: 28622038

arXiv 2018

[2] [2]

Curran, Allyson Gallant, Helen Wong, Ca ther- ine Johnson, Alannah Delahunty-Pike, Lynora Saxinger, Derek Chu, Jeannette Comeau, Trudy Flynn, Julie Clegg, and Christopher Dye

Leah Boulos, Janet A. Curran, Allyson Gallant, Helen Wong, Ca ther- ine Johnson, Alannah Delahunty-Pike, Lynora Saxinger, Derek Chu, Jeannette Comeau, Trudy Flynn, Julie Clegg, and Christopher Dye

work page

[3] [3]

Salim Haﬁd, Sebastian Schellhammer, Sandra Bringay, Konstantin Todorov, and Stefan Dietze. 2022. SciTweets - A Dataset and An- notation Framework for Detecting Scientiﬁc Online Discourse. In Proceedings of the 31st ACM International Conference on Inf ormation & Knowledge Management (Atlanta, GA, USA) (CIKM ’22). Associ- ation for Computing Machinery, New ...

arXiv 2022

[4] [4]

Beizhe Hu, Qiang Sheng, Juan Cao, Yuhui Shi, Yang Li, Dand- ing Wang, and Peng Qi. 2023. Bad Actor, Good Advisor: Explor- ing the Role of Large Language Models in Fake News Detection. arXiv: 2309.12247 [cs.CL]

arXiv 2023

[5] [5]

Yue Huang and Lichao Sun. 2023. Harnessing the Power of ChatGPT in Fake News: An In-Depth Exploration in Generation, Detection and Explanation. arXiv: 2310.05046 [cs.CL]

arXiv 2023

[6] [6]

Rohit Kumar Kaliyar, Anurag Goswami, and Pratik Narang. 2021. FakeBERT: Fake news detection in social media with a BERT-based deep learning approach. Multimedia Tools and Applications 80, 8 (Jan. 2021), 11765–11788. h/t_tps://doi.org/10.1007/s11042-020-10183-2

doi:10.1007/s11042-020-10183-2 2021

[7] [7]

Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, Sebastian Riedel, and Douwe Kiela

[8] [8]

Cheng Li, Jindong Wang, Yixuan Zhang, Kaijie Zhu, Wenxin Hou, Jianxun Lian, Fang Luo, Qiang Yang, and Xing Xie. 2023. Large Lan- guage Models Understand and Can be Enhanced by Emotional Stimuli. arXiv: 2307.11760 [cs.CL]

arXiv 2023