Elamur Research⌕
☰
home / papers / 2026 / “Decode the Law”: Towards Legal Text Simplification with Large Language Models

“Decode the Law”: Towards Legal Text Simplification with Large Language Models

Summary

Legal documents are hard for non-specialists to read, yet resources for simplifying legal language automatically are scarce. The authors build SIMPLE-LAW, a semi-synthetic parallel corpus of 6,230 legal passages paired with plain-language rewrites. The source passages come from public Indian, US and UK legal portals and include statutes, judicial opinions, notices and contract clauses. GPT-3.5-Turbo produced the simplified versions using 1-shot, 2-shot and chain-of-thought prompts. The prompt demonstrations were drawn from a small gold set written by two legal experts. Readability formulas confirm that the rewrites are easier to read than the originals. The authors also checked semantic preservation and entailment, and ran a human evaluation. They then benchmarked Unsloth builds of Mistral, LLaMA-3.2, Gemma and Qwen-2.5, comparing in-context learning with parameter-efficient fine-tuning. Evaluation used BERTScore, ROUGE, SARI and a hallucination score. Fine-tuned models clearly beat prompted ones on both simplification quality and factual consistency. The work supplies a public multi-jurisdiction benchmark and baselines for making legal text accessible.

Key points

  • Corpus composition: SIMPLE-LAW has 6,230 aligned pairs (legal_text → simplified_text).
    • About 3,000+ pairs are Indian, drawn from sources such as indiacode.nic.in and Lighteval legal_summarization.
    • About 1,500+ are US and about 1,500+ are UK.
  • Length reduction: Simplification roughly halves passage length, from about 292.7 to about 137.2 words on average.
  • Readability gains (original → simplified):
    • Flesch Reading Ease improves from 16.34 to about 32.3.
    • FKGL drops from 23.56 to 15.08.
    • SMOG drops from about 16.97 to 13.99.
    • Dale–Chall falls only slightly, from 9.83 to 9.64.
    • Table 1 also reports BillSum for comparison.
  • Models tested: Unsloth checkpoints, including:
    • Mistral-7B-Instruct v0.3
    • LLaMA-3.2 1B/3B
    • Gemma-3-4b-it
    • Qwen2.5 1.5B/7B
  • Availability: The dataset and code are released on GitHub (mohammeddanishrabbani/Legal-Simplification-mrabbani).

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

Fields: Linguistics & languages · Machine translation · Law

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/2i7en3xycmhi

@inproceedings{rabbani2026decode,
  title = {{“Decode the Law”: Towards Legal Text Simplification with Large Language Models}},
  author = {Rabbani, Mohammed Danish and Roy, Subhadeep and Mitra, Sayantan and Saha, Tulika},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/2i7en3xycmhi},
  url = {https://doi.org/10.63317/2i7en3xycmhi}
}

References

53 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Mistral AI. 2023. Mistral 7b. https://mistral.ai/news/announcing-mistral-7b. Accessed: 2025-05-19. ↗
  2. Suha S. Al-Thanyyan and Aqil M. Azmi. 2021. Automated text simplification: A survey. ACM Comput. Surv., 54(2). find
  3. Fernando Alva-Manchego, Louis Martin, Antoine Bordes, Carolina Scarton, Benoît Sagot, and Lucia Specia. 2020a. ASSET: A dataset for tuning and evaluation of sentence simplification models with multiple rewriting transformations. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 4668–4679, Online. Association for Computational Linguistics. find our card
  4. Fernando Alva-Manchego, Louis Martin, Carolina Scarton, and Lucia Specia. 2020b. Asset: A dataset for tuning and evaluation of sentence simplification models with multiple rewriting transformations. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 4668–4679. find our card
  5. Mark Athugodage, Olga Mitrofanove, and Vadim Gudkov. 2024. Transfer learning for Russian legal text simplification. In Proceedings of the 3rd Workshop on Tools and Resources for People with REAding DIfficulties (READI) @ LREC-COLING 2024, pages 59–69, Torino, Italia. ELRA and ICCL. find
  6. Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter, Christopher Hesse, Mark Chen, Eric Sigler, Mateusz Litwin, Scott Gray, Benjamin Chess, Jack Clark, Christopher Berner, Sam McCandlish, Alec Radford, Ilya Sutskever, and Dario Amodei. 2020. Language models are few-shot learners. ↗
  7. Mert Cemri, Tolga Çukur, and Aykut Koç. 2022. Unsupervised simplification of legal texts. ↗
  8. Jeanne S. Chall and Edgar Dale. 1995. Readability Revisited: The New Dale-Chall Readability Formula. Brookline Books, Cambridge, MA. find
  9. Yinbo Chen and Xiaolong Wang. 2022. Transformers as meta-learners for implicit neural representations. find
  10. Google DeepMind. 2024. Gemma: Open models based on gemini research and technology. https://ai.google.dev/gemma. Accessed: 2025-05-19. ↗
  11. Tim Dettmers, Mike Lewis, Sam Shleifer, and Luke Zettlemoyer. 2022. 8-bit optimizers via block-wise quantization. ↗ our card
  12. Tim Dettmers, Artidoro Pagnoni, Ari Holtzman, and Luke Zettlemoyer. 2023. Qlora: Efficient finetuning of quantized llms. arXiv preprint arXiv:2305.14314. ↗
  13. Anand Devaraj and et al. 2023. Medsimplify: A dataset for simplification of medical texts. In Proceedings of ACL BioNLP. find
  14. European Commission. n.d. Clear writing in the european commission. Accessed: 2025-05-18. find
  15. Rudolf Flesch. 1948. A new readability yardstick. Journal of Applied Psychology, 32(3):221–233. ↗
  16. Isabel Gallegos. 2022. The right to remain plain: Summarization and simplification of legal documents. find
  17. Aparna Garimella, Abhilasha Sancheti, Vinay Aggarwal, Ananya Ganesh, Niyati Chhaya, and Nandakishore Kambhatla. 2022. Text simplification for legal domain: Insights and challenges. In Proceedings of the Natural Legal Language Processing Workshop 2022, pages 296–304, Abu Dhabi, United Arab Emirates (Hybrid). Association for Computational Linguistics. find
  18. John Gibbons. 2003. Forensic Linguistics: An Introduction to Language in the Justice System. Blackwell Publishing, Oxford, UK. find
  19. Aaron Grattafiori and et al. 2024. Llama 3 herd of models. https://arxiv.org/abs/2407.21783. Accessed: 2025-05-19. ↗
  20. D. Han, M. Han, and the Unsloth Team. 2023. Unsloth. https://github.com/unslothai/unsloth. ↗
  21. Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2021. Lora: Low-rank adaptation of large language models. find
  22. J. Peter Kincaid, Robert P. Fishburne, Richard L. Rogers, and Brad S. Chissom. 1975. Derivation of new readability formulas (automated readability index, fog count and flesch reading ease formula) for navy enlisted personnel. Technical Report Research Branch Report 8-75, Naval Air Station Memphis, Chief of Naval Technical Training. ↗
  23. Anastassia Kornilova and Vladimir Eidelman. 2019. Billsum: A corpus for automatic summarization of us legislation. In Proceedings of the 2nd Workshop on New Frontiers in Summarization, pages 48–56. find
  24. Nathan Lambert. 2025. Reinforcement learning from human feedback. ↗
  25. Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Ves Stoyanov, and Luke Zettlemoyer. 2019. Bart: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension. find our card
  26. Lighteval. 2023. legal_summarization. find
  27. Chin-Yew Lin. 2004. ROUGE: A package for automatic evaluation of summaries. In Text Summarization Branches Out, pages 74–81, Barcelona, Spain. Association for Computational Linguistics. find
  28. Aman Madaan and et al. 2022. Text simplification with human-in-the-loop feedback. In ACL. find
  29. Louis Martin, Benoît Sagot, Éric de la Clergerie, and Antoine Bordes. 2020. Controllable sentence simplification. ↗
  30. G. Harry McLaughlin. 1969. Smog grading: A new readability formula. Journal of Reading, 12(8):639–646. find
  31. Junya Nishihara and et al. 2022. Readability-controllable text simplification with enhanced reward mechanism. In EMNLP. find
  32. OpenAI. 2023. Gpt-3.5-turbo technical report. https://platform.openai.com/docs/models/gpt-3.5-turbo. Accessed: 2025-05-23. ↗
  33. PWOnlyIAS. 2025. Simplifying legal language. https://pwonlyias.com/editorial-analysis/ simplifying-legal-language/. Accessed: 2025-05-09. ↗
  34. Nils Reimers. 2021. all-minilm-l6-v2. https://huggingface.co/sentence-transformers/ all-MiniLM-L6-v2. Accessed: 2025-05-23. ↗
  35. Nils Reimers and Iryna Gurevych. 2019. Sentence-BERT: Sentence embeddings using Siamese BERT-networks. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 3982–3992, Hong Kong, China. Association for Computational Linguistics. ↗
  36. Aditya G. Saputra. 2025. Medical diagnosis chatbot assistant. https://medium.com/@adityagofi/ medical-diagnosis-chatbot-assistant-71fb0d33dce Advaith Siddharthan. 2014. A survey of research on text simplification. International Journal of Applied Linguistics, 165(2):259–298. ↗
  37. Qwen Team. 2024. Qwen2.5: A party of foundation models! https://qwenlm.github.io/blog/qwen2.5/. Accessed: 2025-05-19. ↗
  38. Don Tuggener, Pius von Däniken, Mark Cieliebak, and Hans Hofmann. 2020. Ledgar: A large-scale multi-label dataset for document classification in the legal domain. In Proceedings of the 12th Language Resources and Evaluation Conference, pages 1235–1241. find
  39. Utkarsh Ujwal, Sai Sri Harsha Surampudi, Sayantan Mitra, and Tulika Saha. 2024. "reasoning before responding": Towards legal long-form question answering with interpretability. In Proceedings of the 33rd ACM International Conference on Information and Knowledge Management, CIKM ’24, page 4922–4930, New York, NY, USA. Association for Computing Machinery. find
  40. United Nations Development Programme. 2022. Access to justice. https://www.undp.org/justice/access-to-justice. Accessed: 2025-05-18. ↗
  41. Unsloth. 2024. unsloth/gemma-3-4b-it. https://huggingface.co/unsloth/gemma-3-4b-it. Accessed: 2025-05-15. ↗
  42. Unsloth. 2025a. unsloth/llama-3.2-1b-instruct-bnb-4bit. https://huggingface.co/unsloth/Llama-3.2-1B-Instruct-bnb-4bit. Accessed: 2025-05-16. ↗
  43. Unsloth. 2025b. unsloth/llama-3.2-3b. https://huggingface.co/unsloth/Llama-3.2-3B. Accessed: 2025-05-16. ↗
  44. Unsloth. 2025c. unsloth/mistral-7b-instruct-v0.3-bnb-4bit. https://huggingface.co/unsloth/mistral-7b-instruct-v0. 3-bnb-4bit. Accessed: 2025-05-16. ↗
  45. Unsloth. 2025d. unsloth/qwen2.5-1.5b-instruct-bnb-4bit. https://huggingface.co/unsloth/Qwen2.5-1.5B-Instruct-bnb-4bit. Accessed: 2025-05-16. ↗
  46. Unsloth. 2025e. unsloth/qwen2.5-7b. https://huggingface.co/unsloth/Qwen2.5-7B. Accessed: 2025-05-16. ↗
  47. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, et al. 2022. Chain-of-thought prompting elicits reasoning in large language models. arXiv preprint arXiv:2201.11903. ↗ our card
  48. Adina Williams, Nikita Nangia, and Samuel Bowman. 2018. A broad-coverage challenge corpus for sentence understanding through inference. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers), pages 1112– 1122, New Orleans, Louisiana. Association for Computational Linguistics. ↗
  49. Wei Xu, Chris Callison-Burch, and Courtney Napoles. 2015. Problems in current text simplification research: New data can help. In Proceedings of NAACL. ↗ our card
  50. Wei Xu, Courtney Napoles, Ellie Pavlick, Quanze Chen, and Chris Callison-Burch. 2016. Optimizing statistical machine translation for text simplification. Transactions of the Association for Computational Linguistics, 4:401–415. ↗ our card
  51. Wenpeng Yin, Jamaal Hay, and Dan Roth. 2019. Benchmarking zero-shot text classification: Datasets, evaluation and entailment approach. ↗ our card
  52. Tianyi Zhang, Varsha Kishore, Felix Wu, Kilian Q. Weinberger, and Yoav Artzi. 2020. Bertscore: Evaluating text generation with bert. ↗
  53. Xinyu Zhang and Mirella Lapata. 2017. Sentence simplification with deep reinforcement learning. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 584–594. find our card

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/rabbani-2026-decode-law-legal-text-simplification/