Elamur Research⌕
☰
home / papers / 2025 / LLM-based NLG Evaluation: Current Status and Challenges

LLM-based NLG Evaluation: Current Status and Challenges

Paper · 2025 · Mingqi Gao, Xinyu Hu, Xunjian Yin et al.

Summary

Judging natural language generation is difficult because a good output has no single correct form and must satisfy several quality criteria at once. Overlap metrics like BLEU, and even model-based scores like BARTScore, agree poorly with human judgments and are easy to fool. This Peking University survey reviews roughly 100 studies, collected from Google Scholar up to June 2024 and including many arXiv preprints, that use LLMs with more than one billion parameters as evaluators of tasks such as translation, summarization, dialogue and data-to-text. The authors group the work into four families: scores computed from LLM embeddings or token probabilities, prompting existing LLMs, fine-tuning LLMs on labelled evaluation data, and evaluation shared between humans and LLMs. For each family they weigh robustness, efficiency and bias. A side-by-side comparison on SummEval shows that the strongest LLM-based judges clearly outperform classic metrics, with cost rising alongside quality. The survey ends by calling for unified benchmarks, research beyond English, and new forms of human–LLM collaboration. These gaps matter directly for building evaluators for Ukrainian and other under-resourced languages.

Key points

  • Taxonomy and example methods:
    • LLM-derived metrics: GPTScore sums token log-probabilities of the output given an instruction and context. FFLM and DELTA-SCORE measure how probabilities shift under changed conditions. Es et al. compute embedding similarity with text-embedding-ada-002.
    • Prompting: G-Eval.
    • Fine-tuning: INSTRUCTSCORE, Prometheus 2, Themis, TIGERScore, CompassJudger-1.
    • Human–LLM collaboration: InteractEval.
  • SummEval correlations (overall score):
    • BERTScore 0.231 and BARTScore 0.329.
    • GPTScore (Qwen-2.5, 72B) 0.436.
    • G-Eval (GPT-4) 0.523 and Themis (8B) 0.553.
    • InteractEval (GPT-4) reaches about 0.71–0.73, the best result.
  • Cost ranking, from cheapest to most expensive:
    • LLM-derived metrics and prompting open-source models cost about the same.
    • Fine-tuning costs roughly as much as prompting proprietary APIs.
    • Human–LLM collaboration is the most expensive.
  • Weaknesses of LLM-derived metrics:
    • Stress tests (He et al. 2023) expose blind spots.
    • Closed models do not expose logits, so these metrics cannot be applied to them.
    • Model-based metrics carry more social bias across six sensitive attributes (Sun et al. 2022).
  • Gaps for future work:
    • Almost all research targets English, and authors urge caution for non-Latin scripts. Fine-tuning on low-resource languages looks promising (Hada et al. 2024).
    • Existing human-judgment sets are small, cover a single task, and contain outputs from outdated systems such as Pointer Network.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Peking UniversityBeijing · China100 works · 35,618 citations

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 9,086 worksArtificial Intelligence · 8,889 worksArts and Humanities · 9,296 worksLanguage and Linguistics · 9,206 works

🧭 Topics 2Research subjects and languages the work is about.Evaluation & benchmarks · 685 worksText generation · 145 works

Cite

Cite as Mingqi Gao, Xinyu Hu, Xunjian Yin, Jie Ruan, Xiao Pu and Xiaojun Wan. 2025. LLM-based NLG Evaluation: Current Status and Challenges. Computational Linguistics.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.1162/coli_a_00561

@article{gao2025llm,
  title = {{LLM-based NLG Evaluation: Current Status and Challenges}},
  author = {Mingqi Gao, Xinyu Hu, Xunjian Yin, Jie Ruan, Xiao Pu and Xiaojun Wan},
  journal = {Computational Linguistics},
  year = {2025},
  doi = {10.1162/coli\_a\_00561},
  url = {https://doi.org/10.1162/coli_a_00561}
}

References

59 references, 29 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Amodei, Dario, Chris Olah, Jacob Steinhardt, Paul F. Christiano, John Schulman, and Dan Mané. 2016. Concrete problems in AI safety. CoRR, abs/1606.06565.
  2. Anil, Rohan, Andrew M. Dai, Orhan Firat, Melvin Johnson, Dmitry Lepikhin, Alexandre Passos, Siamak Shakeri, Emanuel Taropa, Paige Bailey, Zhifeng Chen, et al. 2023. PaLM 2 technical report. Computing Research Repository, arXiv:2305.10403.
  3. Ashktorab, Zahra, Michael Desmond, Qian Pan, James M. Johnson, Martin Santillan Cooper, Elizabeth M. Daly, Rahul Nair, Tejaswini Pedapati, Swapnaja Achintalwar, and Werner Geyer. 2024. Aligning human and LLM judgments: Insights from EvalAssist on task-specific evaluations and AI-assisted assessment… — [arXiv:2410.00873](https://arxiv.org/abs/2410.00873) find
  4. Bai, Yushi, Jiahao Ying, Yixin Cao, Xin Lv, Yuze He, Xiaozhi Wang, Jifan Yu, Kaisheng Zeng, Yijia Xiao, Haozhe Lyu, Jiayin Zhang, Juanzi Li, and Lei Hou. 2023. Benchmarking foundation models with language-model-as-an-examiner. CoRR, abs/2306.04181. Bavaresco, Anna, Raffaella Bernardi, Leonardo Berto…
  5. Cao, Maosong, Alexander Lam, Haodong Duan, Hongwei Liu, Songyang Zhang, and Kai Chen. 2024. CompassJudger-1: All-in-one judge model helps model evaluation and evolution. arXiv preprint arXiv:2410.16256. — [arXiv:2410.16256](https://arxiv.org/abs/2410.16256) find
  6. Cegin, Jan, Jakub Simko, and Peter Brusilovsky. 2023. ChatGPT to replace crowdsourcing of paraphrases for intent classification: Higher diversity and comparable model robustness. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 1889–1905. https://doi.o… ↗
  7. Chan, Chi-Min, Weize Chen, Yusheng Su, Jianxuan Yu, Wei Xue, Shanghang Zhang, Jie Fu, and Zhiyuan Liu. 2023. ChatEval: Towards better LLM-based evaluators through multi-agent debate. CoRR, abs/2308.07201.
  8. Chang, Yapei, Kyle Lo, Tanya Goyal, and Mohit Iyyer. 2023. BooookScore: A systematic exploration of book-length summarization in the era of LLMs. CoRR, abs/2310.00785. find
  9. Chang, Yupeng, Xu Wang, Jindong Wang, Yuan Wu, Linyi Yang, Kaijie Zhu, Hao Chen, Xiaoyuan Yi, Cunxiang Wang, Yidong Wang, et al. 2024. A survey on evaluation of large language models. ACM Transactions on Intelligent Systems and Technology, 15(3):39:1–39:45. https://doi.org/10.1145/3641289 Chen, Mark… ↗
  10. Chiang, David Cheng-Han and Hung-yi Lee. 2023a. Can large language models be an alternative to human evaluations? In Proceedings of ACL (1), pages 15607–15631. https://doi.org/10.18653/v1/2023.acl-long.870 Chiang, David Cheng-Han and Hung-yi Lee. 2023b. A closer look into using large language model… ↗
  11. Cohen, Roi, May Hamri, Mor Geva, and Amir Globerson. 2023. LM vs LM: Detecting factual errors via cross examination. In Proceedings of EMNLP, pages 12621–12640. https://doi.org/10.18653/v1/2023.emnlp-main.778 Deriu, Jan, Álvaro Rodrigo, Arantxa Otegi, Guillermo Echegoyen, Sophie Rosset, Eneko Agir… ↗
  12. Tixier, and Michalis Vazirgiannis. 2022. FrugalScore: Learning cheaper, lighter and faster evaluation metrics for automatic text generation. In Proceedings of ACL (1), pages 1305–1318. https://doi.org/10.18653/v1/2022 .acl-long.93 Es, Shahul, Jithin James, Luis Espinosa Anke, and Steven Schockaert.… ↗
  13. Fabbri, Alexander R., Wojciech Kryscinski, Bryan McCann, Caiming Xiong, Richard Socher, and Dragomir Radev. 2021. SummEval: Re-evaluating summarization evaluation. Transactions of the Association for Computational Linguistics, 9:391–409. https://doi.org/10.1162/tacl_a_00373 Freitag, Markus, Ricardo… ↗
  14. Fu, Jinlan, See-Kiong Ng, Zhengbao Jiang, and Pengfei Liu. 2023a. GPTScore: Evaluate as you desire. CoRR, abs/2302.04166. find
  15. Fu, Xue-Yong, Md. Tahmid Rahman Laskar, Cheng Chen, and Shashi Bhushan T. N. 2023b. Are large language models reliable judges? A study on the factuality evaluation capabilities of LLMs. CoRR, abs/2311.00681. find
  16. Gao, Mingqi, Jie Ruan, Renliang Sun, Xunjian Yin, Shiping Yang, and Xiaojun Wan. 2023. Human-like summarization evaluation with ChatGPT. CoRR, abs/2304.02554. find
  17. Gehrmann, Sebastian, Tosin P. Adewumi, Karmanya Aggarwal, Pawan Sasanka Ammanamanchi, Aremu Anuoluwapo, Antoine Bosselut, Khyathi Raghavi Chandu, Miruna-Adriana Clinciu, Dipanjan Das, Kaustubh D. Dhole, et al. 2021. The GEM benchmark: Natural language generation, its evaluation and metrics. CoRR, ab… find
  18. Guan, Jian, Jesse Dodge, David Wadden, Minlie Huang, and Hao Peng. 2023. Language models hallucinate, but may excel at fact verification. CoRR, abs/2310.14564. https://doi.org/10.18653/v1/2024.naacl-long.62 Hada, Rishav, Varun Gumma, Mohamed Ahmed, Kalika Bali, and Sunayana Sitaram. 2024. METAL: To… ↗
  19. He, Hangfeng, Hongming Zhang, and Dan Roth. 2023. SocREval: Large language models with the Socratic method for reference-free reasoning evaluation. CoRR, abs/2310.00074. find
  20. He, Tianxing, Jingyu Zhang, Tianle Wang, Sachin Kumar, Kyunghyun Cho, James Glass, and Yulia Tsvetkov. 2023. On the blind spots of model-based evaluation metrics for text generation. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), p… find
  21. Jia, Qi, Siyu Ren, Yizhu Liu, and Kenny Q. Zhu. 2023. Zero-shot faithfulness evaluation for text summarization with foundation language model. In Proceedings of EMNLP, pages 11017–11031. https://doi.org/10.18653/v1/2023.emnlp-main.679 Jiang, Dongfu, Yishan Li, Ge Zhang, Wenhao Huang, Bill Yuchen Li… ↗
  22. Jones, Erik and Jacob Steinhardt. 2022. Capturing failures of large language models via human cognitive biases. In Proceedings of NeurIPS. find
  23. Jung, Jaehun, Lianhui Qin, Sean Welleck, Faeze Brahman, Chandra Bhagavatula, Ronan Le Bras, and Yejin Choi. 2022. Maieutic Prompting: Logically consistent reasoning with recursive explanations. In Proceedings of EMNLP. https://doi.org/10.18653/v1/2022.emnlp -main.82 Ke, Pei, Bosi Wen, Zhuoer Feng,… ↗
  24. Kim, Joonghoon, Saeran Park, Kiyoon Jeong, Sangmin Lee, Seung Hun Han, Jiyoon Lee, and Pilsung Kang. 2023a. Which is better? Exploring prompting strategy for LLM-based metrics. CoRR, abs/2311.03754. https://doi.org/10 .18653/v1/2023.eval4nlp-1.14 Kim, Seungone, Jamin Shin, Yejin Cho, Joel Jang, Shay… find
  25. Kim, Seungone, Juyoung Suk, Ji Yong Cho, Shayne Longpre, Chaeeun Kim, Dongkeun Yoon, Guijin Son, Yejin Choi, Sheikh Shafayat, Jinheon Baek, et al. 2024a. The BiGGen Bench: A principled benchmark for fine-grained evaluation of language models with language models. CoRR, abs/2406.05761. find
  26. Kim, Seungone, Juyoung Suk, Shayne Longpre, Bill Yuchen Lin, Jamin Shin, Sean Welleck, Graham Neubig, Moontae Lee, Kyungjae Lee, and Minjoon Seo. 2024b. Prometheus 2: An open source language model specialized in evaluating other language models. In Proceedings of the 2024 Conference on Empirical Met…
  27. Kotonya, Neema, Saran Krishnasamy, Joel R. Tetreault, and Alejandro Jaimes. 2023. Little giants: Exploring the potential of small LLMs as evaluation metrics in summarization in the Eval4NLP 2023 shared task. CoRR, abs/2311.00686. https://doi.org/10.18653/v1/2023 .eval4nlp-1.17 Kwon, Woosuk, Zhuohan… ↗
  28. Li, Qintong, Leyang Cui, Lingpeng Kong, and Wei Bi. 2023b. Collaborative evaluation: Exploring the synergy of large language models and humans for open-ended generation evaluation. CoRR, abs/2310.19740. — [arXiv:2310.19740](https://arxiv.org/abs/2310.19740) find
  29. Li, Ruosen, Teerth Patel, and Xinya Du. 2023. PRD: Peer rank and discussion improve large language model based evaluations. CoRR, abs/2307.02762. Li, Yujia, David Choi, Junyoung Chung, Nate Kushman, Julian Schrittwieser, Rémi Leblond, Tom Eccles, James Keeling, Felix Gimeno, Agustin Dal Lago, et al… find
  30. Lin, Yen-Ting and Yun-Nung Chen. 2023. LLM-Eval: Unified multi-dimensional automatic evaluation for open-domain conversations with large language models. In NLP4ConvAI 2023. arXiv:2305.13711. https://doi.org/10.18653/v1/2023.nlp4convai-1.5 Liu, Yang, Dan Iter, Yichong Xu, Shuohang Wang, Ruochen Xu,… ↗
  31. Liu, Yuxuan, Tianchi Yang, Shaohan Huang, Zihan Zhang, Haizhen Huang, Furu Wei, Weiwei Deng, Feng Sun, and Qi Zhang. 2023e. Calibrating LLM-based evaluator. CoRR, abs/2309.13308.
  32. Liusie, Adian, Potsawee Manakul, and Mark J. F. Gales. 2023. LLM comparative assessment: Zero-shot NLG evaluation through pairwise comparisons using large language models. Computing Research Repository, arxiv:2307.07889.
  33. Lu, Qingyu, Baopu Qiu, Liang Ding, Liping Xie, and Dacheng Tao. 2023. Error analysis prompting enables human-like translation evaluation in large language models: A case study on ChatGPT. CoRR, abs/2303.13809. https://doi.org/10.20944/preprints202303 .0255.v1 Luo, Zheheng, Qianqian Xie, and Sophia… ↗
  34. Manakul, Potsawee, Adian Liusie, and Mark J. F. Gales. 2023. SelfCheckGPT: Zero-resource black-box hallucination detection for generative large language models. In Proceedings of EMNLP, pages 9004–9017. https://doi.org/10.18653/v1/2023.emnlp-main.557 Mendonça, John, Patrı́cia Pereira, João Paulo… ↗
  35. Menick, Jacob, Maja Trebacz, Vladimir Mikulik, John Aslanides, H. Francis Song, Martin J. Chadwick, Mia Glaese, Susannah Young, Lucy Campbell-Gillingham, Geoffrey Irving, and Nat McAleese. 2022. Teaching language models to support answers with verified quotes. CoRR, abs/2203.11147.
  36. Murugadoss, Bhuvanashree, Christian Poelitz, Ian Drosos, Vu Le, Nick McKenna, Carina Suzana Negreanu, Chris Parnin, and Advait Sarkar. 2024. Evaluating the evaluator: Measuring LLMs’ adherence to task evaluation instructions. arXiv preprint arXiv:2408.08781. https://doi.org /10.1609/aaai.v39i18.3415… ↗
  37. Ni’mah, Iftitahu, Meng Fang, Vlado Menkovski, and Mykola Pechenizkiy. 2023. NLG evaluation metrics beyond correlation analysis: An empirical metric preference checklist. In Proceedings of ACL (1), pages 1240–1266. https://doi.org/10.18653/v1/2023.acl-long.69 Ostyakova, Lidiia, Veronika Smilga, Ksen… ↗
  38. Papineni, Kishore, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. BLEU: a method for automatic evaluation of machine translation. In Proceedings of ACL, pages 311–318. https://doi.org/10.3115/1073083.1073135 Pozdniakov, Stanislav, Jonathan Brazil, Solmaz Abdi, Aneesha Bakharia, Shazia Sadiq, Drag… ↗
  39. Ribeiro, Marco Túlio and Scott M. Lundberg. 2022. Adaptive testing and debugging of NLP models. In Proceedings of ACL (1), pages 3253–3267. https://doi.org/10.18653/v1/2022.acl -long.230 Saha, Swarnadeep, Omer Levy, Asli Celikyilmaz, Mohit Bansal, Jason Weston, and Xian Li. 2023. Branch-solve-merg… ↗
  40. Sai, Ananya B., Tanay Dixit, Dev Yashpal Sheth, Sreyas Mohan, and Mitesh M. Khapra. 2021. Perturbation checklists for evaluating NLG evaluation metrics. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 7219–7234. https://doi.org/10.18653/v1 /2021.emnlp… find
  41. Saunders, William, Catherine Yeh, Jeff Wu, Steven Bills, Long Ouyang, Jonathan Ward, and Jan Leike. 2022. Self-critiquing models for assisting human evaluators. CoRR, abs/2206.05802.
  42. Shankar, Shreya, J.D. Zamfirescu-Pereira, Bjoern Hartmann, Aditya Parameswaran, and Ian Arawjo. 2024. Who validates the validators? Aligning LLM-assisted evaluation of LLM outputs with human preferences. In Proceedings of the 37th Annual ACM Symposium on User Interface Software and Technology, UIST…
  43. Stent, Amanda, Matthew Marge, and Mohit Singhai. 2005. Evaluating evaluation methods for generation in the presence of variation. In CICLing, volume 3406 of Lecture Notes in Computer Science, pages 341–351. https://doi.org/10.1007/978-3-540-30586 -6 38 Sun, Tianxiang, Junliang He, Xipeng Qiu, and X… ↗
  44. Touvron, Hugo, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, et al. 2023. Llama 2: Open foundation and fine-tuned chat models. CoRR, abs/2307.09288. van der Lee, Chris, Albert Gatt, Emiel van Miltenburg,… find
  45. Wang, Jiaan, Yunlong Liang, Fandong Meng, Haoxiang Shi, Zhixu Li, Jinan Xu, Jianfeng Qu, and Jie Zhou. 2023a. Is ChatGPT a good NLG evaluator? A preliminary study. In Proceedings of the 4th New Frontiers in Summarization Workshop, arXiv:2303 .04048. https://doi.org/10.18653/v1/2023.newsum-1.1 Wang,… ↗
  46. Wang, Tianlu, Ilia Kulikov, Olga Golovneva, Ping Yu, Weizhe Yuan, Jane Dwivedi-Yu, Richard Yuanzhe Pang, Maryam Fazel-Zarandi, Jason Weston, and Xian Li. 2024. Self-Taught evaluators. arXiv preprint arXiv:2408.02666.
  47. Wang, Tianlu, Ping Yu, Xiaoqing Ellen Tan, Sean O’Brien, Ramakanth Pasunuru, Jane Dwivedi-Yu, Olga Golovneva, Luke Zettlemoyer, Maryam Fazel-Zarandi, and Asli Celikyilmaz. 2023c. Shepherd: A critic for language model generation. CoRR, abs/2308.04592.
  48. Wang, Yaqing, Jiepu Jiang, Mingyang Zhang, Cheng Li, Yi Liang, Qiaozhu Mei, and Michael Bendersky. 2023d. Automated evaluation of personalized text generation using large language models. CoRR, abs/2310.11593. — [arXiv:2310.11593](https://arxiv.org/abs/2310.11593) find
  49. Wang, Yidong, Zhuohao Yu, Zhengran Zeng, Linyi Yang, Cunxiang Wang, Hao Chen, Chaoya Jiang, Rui Xie, Jindong Wang, Xing Xie, Wei Ye, Shikun Zhang, and Yue Zhang. 2023e. PandaLM: An automatic evaluation benchmark for LLM instruction tuning optimization. CoRR, abs/2306.05087.
  50. Wang, Zifan, Kotaro Funakoshi, and Manabu Okumura. 2023. Automatic answerability evaluation for question generation. CoRR, abs/2309.12546. — [arXiv:2309.12546](https://arxiv.org/abs/2309.12546) find
  51. Wiseman, Sam, Stuart M. Shieber, and Alexander M. Rush. 2017. Challenges in data-to-document generation. In EMNLP, pages 2253–2263. https://doi.org/10 .18653/v1/D17-1239 Wu, Minghao and Alham Fikri Aji. 2023. Style over substance: Evaluation biases for large language models. CoRR, abs/2307.03025.
  52. Wu, Ning, Ming Gong, Linjun Shou, Shining Liang, and Daxin Jiang. 2023. Large language models are diverse role-players for summarization evaluation. In Proceedings of NLPCC (1), pages 695–707. https://doi.org/10.1007/978-3 -031-44693-1 54 Xiao, Ziang, Susu Zhang, Vivian Lai, and Q. Vera Liao. 2023.… ↗
  53. Ye, Xi, Srinivasan Iyer, Asli Celikyilmaz, Veselin Stoyanov, Greg Durrett, and Ramakanth Pasunuru. 2023b. Complementary explanations for effective in-context learning. In Proceedings of ACL (Findings), pages 4469–4484. https://doi.org/10.18653/v1/2023.findings-acl.273 Yin, Kayo and Graham Neubig. 2… ↗
  54. Zhang, Chen, Luis Fernando D’Haro, Yiming Chen, Malu Zhang, and Haizhou Li. 2023a. A comprehensive analysis of the effectiveness of large language models as automatic dialogue evaluators. CoRR, abs/2312.15407. — [arXiv:2312.15407](https://arxiv.org/abs/2312.15407) find
  55. Zhang, Susan, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen, Christopher Dewan, Mona T. Diab, Xian Li, Xi Victoria Lin, et al. 2022. OPT: Open pre-trained transformer language models. CoRR, abs/2205.01068.
  56. Zhang, Xinghua, Bowen Yu, Haiyang Yu, Yangyu Lv, Tingwen Liu, Fei Huang, Hongbo Xu, and Yongbin Li. 2023b. Wider and deeper LLM networks are fairer LLM evaluators. CoRR, abs/2308.01862.
  57. Zhang, Yangjun, Pengjie Ren, and Maarten de Rijke. 2021. A human-machine collaborative framework for evaluating malevolence in dialogues. In Proceedings of ACL/IJCNLP (1), pages 5612–5623. https://doi.org/10.18653/v1/2021.acl-long.436 Zhang, Ying, Stephan Vogel, and Alex Waibel. 2004. Interpreting… ↗
  58. Zheng, Lianmin, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric P. Xing, Hao Zhang, Joseph E. Gonzalez, and Ion Stoica. 2023. Judging LLM-as-a-judge with MT-bench and Chatbot Arena. CoRR, abs/2306.05685. Zhou, Kaitlyn, Su Lin Blodgett, Ada… find
  59. Zhuang, Ziyu, Qiguang Chen, Longxuan Ma, Mingda Li, Yi Han, Yushan Qian, Haopeng Bai, Zixian Feng, Weinan Zhang, and Ting Liu. 2023. Through the lens of core competency: Survey on evaluation of large language models. CoRR, abs/2308.07902. Zouhar, Vilém, Tom Kocmi, and Mrinmaya Sachan. 2025. AI-assi… find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/gao-2025-llm-based-nlg-evaluation-current/