Elamur Research⌕
☰
home / papers / 2026 / Less is More: Knowledge-Aware Compression for Long Legal Judgment Prediction

Less is More: Knowledge-Aware Compression for Long Legal Judgment Prediction

Paper · 2026 · Fanghao Lou, Qiqi Wang, Guanyu Chen et al.

Summary

Predicting court outcomes from long case descriptions is costly for LLMs because attention cost grows quadratically with input length, and crucial legal facts can get lost in lengthy contexts. Existing remedies, such as sparse or hierarchical architectures and extractive or abstractive summarization, either miss scattered dependencies, ignore legal logic, or risk hallucination at high cost. The authors propose KAC, a framework that uses legal knowledge to decide what to keep.

KAC works in four steps:

  • It splits case facts into sentence-level units and embeds them with an LLM.
  • A small MLP picks the most likely statutory provisions, which are paired with their official interpretations (e.g., from the United States Code or China's Criminal Law).
  • Each sentence is scored by its maximum cosine similarity to that knowledge, the scores are normalized, and an Otsu-style adaptive threshold keeps only high-relevance sentences.
  • The compressed text is passed to a fine-tuned LLM for prediction.

Tested on four multilingual, cross-jurisdiction datasets, KAC beats prior methods in both accuracy and efficiency. This makes long-document legal AI more practical, though the gains are small for short texts.

Key points

  • Evaluated on SCOTUS and ECTHR-B (English), SJP (German), and MultiLJP (Chinese), using Qwen2.5-3B and Qwen2-7B backbones.
  • Compression runs in O(n + l²) time, compared with O(n² + m²) for abstractive summarization. The authors also offer an information-theoretic justification showing that KAC maximizes the information shared between the shortened text and the legal knowledge.
  • Otsu thresholding matches the KSW alternative in accuracy while using about 28% less GPU memory and running about 60% faster.
  • Top-3 provision selection is the default: Top-1 and Top-2 perform worse, results level off above 3, and cost rises with larger M.
  • Code is released at github.com/Statistical-NLP-Lab/KAC. The authors caution that the system is a research tool, not a substitute for judges.

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

🎓Harbin Institute of TechnologyHarbin · China158 works · 28,841 citations🎓University of AucklandAuckland · New Zealand11 works · 888 citations🎓Nankai UniversityTianjin · China24 works · 762 citations

 Fields, topics and concepts

📚 Shelves 7The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksSocial Sciences · 1,772 worksPolitical Science and International Relations · 182 worksLaw · 77 works

🧭 Topics 2Research subjects and languages the work is about.Efficient models & compression · 242 worksLegal AI · 134 works

Cite

Cite as Fanghao Lou, Qiqi Wang, Guanyu Chen, Senbo Zhang, Kaiqi Zhao, Qian Liu and Huijia Li. 2026. Less is More: Knowledge-Aware Compression for Long Legal Judgment Prediction. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-acl.1450

@inproceedings{lou2026less,
  title = {{Less is More: Knowledge-Aware Compression for Long Legal Judgment Prediction}},
  author = {Fanghao Lou, Qiqi Wang, Guanyu Chen, Senbo Zhang, Kaiqi Zhao, Qian Liu and Huijia Li},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.1450},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.1450}
}

References

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

  1. Joshua Ainslie, Santiago Ontañón, Chris Alberti, Vaclav Cvicek, Zachary Fisher, Philip Pham, Anirudh Ravula, Sumit Sanghai, Qifan Wang, and Li Yang. 2020. ETC: Encoding long and structured inputs in transformers. In EMNLP, pages 268–284. ↗
  2. Saeed Amizadeh, Sara Abdali, Yinheng Li, and Kazuhito Koishida. 2025. Hierarchical self-attention: Generalizing neural attention mechanics to multi-scale problems. In NIPS. find
  3. Jinze Bai, Shuai Bai, Yunfei Chu, Zeyu Cui, Kai Dang, Xiaodong Deng, Yang Fan, Wenbin Ge, Yu Han, Fei Huang, Binyuan Hui, Luo Ji, Mei Li, Junyang Lin, Runji Lin, Dayiheng Liu, Gao Liu, Chengqiang Lu, Keming Lu, and 29 others. 2023. Qwen technical report.
  4. Iz Beltagy, Matthew E. Peters, and Arman Cohan. 2020. Longformer: The long-document transformer. Preprint, arXiv:2004.05150. ↗
  5. Jesús Calleja, Thierry Etchegoyhen, and David Ponce. 2024. Automating easy read text segmentation. In Findings of EMNLP 2024, pages 11876–11894. find
  6. Ilias Chalkidis, Manos Fergadiotis, Dimitrios Tsarapatsanis, Nikolaos Aletras, Ion Androutsopoulos, and Prodromos Malakasiotis. 2021. Paragraph-level rationale extraction through regularization: A case study on European court of human rights cases. In ACL 2021, pages 226–241. ↗
  7. Ilias Chalkidis, Abhik Jana, Dirk Hartung, Michael Bommarito, Ion Androutsopoulos, Daniel Katz, and Nikolaos Aletras. 2022a. LexGLUE: A benchmark dataset for legal language understanding in English. In ACL 2022, pages 4310–4330. ↗
  8. Ilias Chalkidis, Abhik Jana, Dirk Hartung, Michael Bommarito, Ion Androutsopoulos, Daniel Martin Katz, and Nikolaos Aletras. 2022b. Lexglue: A benchmark dataset for legal language understanding in english. In ACL 2022. ↗
  9. Yapei Chang, Kyle Lo, Tanya Goyal, and Mohit Iyyer. 2024. Booookscore: A systematic exploration of book-length summarization in the era of llms. In ICLR 2024. ↗
  10. Shijing Chen, Mohamed Reda Bouadjenek, Usman Naseem, Basem Suleiman, Shoaib Jameel, Flora Salim, Hakim Hacid, and Imran Razzak. 2025. Leveraging taxonomy and LLMs for improved multimodal hierarchical classification. In COLING 2025, pages 6244–6254. ↗
  11. Rewon Child, Scott Gray, Alec Radford, and Ilya Sutskever. 2019. Generating long sequences with sparse transformers. CoRR, abs/1904.10509. ↗
  12. Yu-Neng Chuang, Tianwei Xing, Chia-Yuan Chang, Zirui Liu, Xun Chen, and Xia Hu. 2024. Learning to compress prompt in natural language formats. In ACL 2024, pages 7756–7767. ↗
  13. DeepSeek-AI. 2025. Deepseek-v3.2-exp: Boosting long-context efficiency with deepseek sparse attention. find
  14. Chenlong Deng, Kelong Mao, Yuyao Zhang, and Zhicheng Dou. 2024. Enabling discriminative reasoning in LLMs for legal judgment prediction. In EMNLP. find
  15. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of deep bidirectional transformers for language understanding. In NAACL-HLT, pages 4171–4186.
  16. Daniele Giofré and Sneha Ghantasala. 2023. Legal-hnet: Mixing legal long-context tokens with hartley transform. Preprint, arXiv:2311.05089. ↗
  17. Neel Guha, Julian Nyarko, Daniel E. Ho, Christopher Ré, Adam Chilton, K. Aditya, Alex Chohlas-Wood, Austin Peters, Brandon Waldon, Daniel N. Rockmore, Diego Zambrano, Dmitry Talisman, Enam Hoque, Faiz Surani, Frank Fagan, Galit Sarfaty, Gregory M. Dickinson, Haggai Porat, Jason Hegland, and 21 others. 2023. Legalbench: A collaboratively built benchmark for measuring legal reasoning in large language models. In NeurIPS.
  18. Pengcheng He, Xiaodong Liu, Jianfeng Gao, and Weizhu Chen. 2021. Deberta: Decoding-enhanced bert with disentangled attention. Preprint, arXiv:2006.03654. ↗
  19. Namgyu Ho, Laura Schmid, and Se-Young Yun. 2023. Large language models are reasoning teachers. In ACL 2023, pages 14852–14882.
  20. Jinpeng Li, Hang Yu, Ziqi Ma, and Peng Qi. 2026. Discovering latent facts from context to construct richer open knowledge graphs. Proceedings of the AAAI Conference on Artificial Intelligence, 40(23):19233– 19241. find
  21. Pengbo Li, Hang Yu, and Xiangfeng Luo. 2025. Context-aware graph neural network for graph-based fraud detection with extremely limited labels. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 12112–12120. find
  22. Ziguang Li, Chao Huang, Xuliang Wang, Haibo Hu, Cole Wyeth, Dongbo Bu, Quan Yu, Wen Gao, Xingwu Liu, and Ming Li. 2024. Lossless data compression by large models. NAT MACH INTELL, 7:794 – 799. ↗
  23. Edward J Hu, yelong shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2022. LoRA: Low-rank adaptation of large language models. In ICLR.
  24. Nayu Liu, Junnan Zhu, Yiming Ma, Zhicong Lu, Wenlei Xu, Yong Yang, Jiang Zhong, and Kaiwen Wei. 2025a. SARA: Salience-aware reinforced adaptive decoding for large language models in abstractive summarization. In ACL 2025. find
  25. Quzhe Huang, Mingxu Tao, Chen Zhang, Zhenwei An, Cong Jiang, Zhibin Chen, Zirui Wu, and Yansong Feng. 2023. Lawyer llama technical report. Preprint, arXiv:2305.15062. ↗
  26. Qian Liu, Rui Mao, Xiubo Geng, and Erik Cambria. 2023. Semantic matching in machine reading comprehension: An empirical study. Information Processing & Management, 60(2):103145. ↗
  27. Ganesh Jawahar, Benoît Sagot, and Djamé Seddah. 2019. What does BERT learn about the structure of language? In ACL 2019, pages 3651–3657.
  28. Qian Liu, Hang Yu, Qiqi Wang, Qi Xu, Jinpeng Li, Zhuoqun Zou, Rui Mao, and Erik Cambria. 2025b. Legal knowledge infusion for large language models: A survey. Information Fusion, page 103426. ↗
  29. Huiqiang Jiang, Qianhui Wu, Chin-Yew Lin, Yuqing Yang, and Lili Qiu. 2023. LLMLingua: Compressing prompts for accelerated inference of large language models. In EMNLP 2023, pages 13358–13376.
  30. Huiqiang Jiang, Qianhui Wu, Xufang Luo, Dongsheng Li, Chin-Yew Lin, Yuqing Yang, and Lili Qiu. 2024. LongLLMLingua: Accelerating and enhancing LLMs in long context scenarios via prompt compression. In ACL 2024, pages 1658–1677. ↗
  31. J.N. Kapur, P.K. Sahoo, and A.K.C. Wong. 1985. A new method for gray-level picture thresholding using the entropy of the histogram. CVGIP, 29(3):273–285. ↗
  32. Daniel Martin Katz, Michael J. Bommarito, II, and Josh Blackman. 2017. A general approach for predicting the behavior of the supreme court of the united states. PLOS ONE, 12(4):1–18. ↗
  33. Strahinja Klem and Noura Al Moubayed. 2025. Llms for llms: A structured prompting methodology for long legal documents. Preprint, arXiv:2509.02241. ↗
  34. Kangil Lee, Jinwoo Jang, Youngjin Lim, and Minsu Shin. 2025. Chain of knowledge graph: Information-preserving multi-document summarization for noisy documents. In Proc. of NeusymBridge @ COLING 2025, pages 1–5. find
  35. James Lee-Thorp, Joshua Ainslie, Ilya Eckstein, and Santiago Ontañón. 2022. FNet: Mixing tokens with Fourier transforms. In NAACL-HLT 2022, pages 4296–4313. ↗
  36. Fanghao Lou, Qiqi Wang, Guanyu Chen, Kaiqi Zhao, and Huijia Li. 2026. Zipljp: Zipped information processor for legal judgment prediction. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 40, pages 32302–32310. find
  37. Yougang Lyu, Jitai Hao, Zihan Wang, Kai Zhao, Shen Gao, Pengjie Ren, Zhumin Chen, Fang Wang, and Zhaochun Ren. 2023. Multi-defendant legal judgment prediction via hierarchical reasoning. In EMNLP 2023, pages 2198–2209. find
  38. Aishik Nagar, Yutong Liu, Andy T. Liu, Viktor Schlegel, Vijay Prakash Dwivedi, Arun-Kumar Kaliya-Perumal, Guna Pratheep Kalanchiam, Yili Tang, and Robby T. Tan. 2025. uMedSum: A unified framework for clinical abstractive summarization. In ACL 2025, pages 2653–2672. find
  39. Shubham Kumar Nigam, Deepak Patnaik Balaramamahanthi, Shivam Mishra, Noel Shallum, Kripabandhu Ghosh, and Arnab Bhattacharya. 2025. NyayaAnumana and INLegalLlama: The largest Indian legal judgment prediction dataset and specialized language model for enhanced decision analysis. In COLING 2025, pages 11135–11160. find
  40. Joel Niklaus, Ilias Chalkidis, and Matthias Stürmer. 2021. Swiss-judgment-prediction: A multilingual legal judgment prediction benchmark. In NLLP 2021.
  41. OpenAI, Josh Achiam, and Steven Adler et al. 2024. Gpt-4 technical report. Preprint, arXiv:2303.08774. ↗
  42. Nobuyuki Otsu. 1979. A threshold selection method from gray-level histograms. IEEE SMC, 9(1):62–66. ↗
  43. Zhuoshi Pan, Qianhui Wu, Huiqiang Jiang, Menglin Xia, Xufang Luo, Jue Zhang, Qingwei Lin, Victor Ruhle, Yuqing Yang, Chin-Yew Lin, H. Vicky Zhao, Lili Qiu, and Dongmei Zhang. 2024. LLMLingua-2: Data distillation for efficient and faithful task-agnostic prompt compression. In Findings of ACL 2024, pages 963–981. ↗
  44. Raghavendra Pappagari, Piotr Zelasko, Jesús Villalba, Yishay Carmiel, and Najim Dehak. 2019. Hierarchical transformers for long document classification. CoRR, abs/1910.10781. ↗
  45. Nishchal Prasad, Mohand Boughanem, and Taoufiq Dkaki. 2024. Exploring large language models and hierarchical frameworks for classification of large unstructured legal documents. In Adv Inf Retr, pages 221–237. find
  46. Qwen, :, An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, Huan Lin, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jingren Zhou, and 25 others. 2025. Qwen2.5 technical report. Preprint, arXiv:2412.15115. ↗
  47. Christophe Rodrigues, Marius Ortega, Aurélien Bossard, and Nédra Mellouli. 2025. Redire: Extreme reduction dimension for extractive summarization. DATA KNOWL ENG, 157:102407. ↗
  48. Jeffrey A. Segal. 1984. Predicting supreme court cases probabilistically: The search and seizure cases, 1962-1981. Am. Polit. Sci. Rev., 78(4):891–900. ↗
  49. Oscar Skean, Md Rifat Arefin, Dan Zhao, Niket Nikul Patel, Jalal Naghiyev, Yann Lecun, and Ravid Shwartz-Ziv. 2025. Layer by layer: Uncovering hidden representations in language models. In ICML 2025, volume 267, pages 55854–55875. ↗
  50. Youwei Song, Jiahai Wang, Zhiwei Liang, Zhiyue Liu, and Tao Jiang. 2020. Utilizing BERT intermediate layers for aspect based sentiment analysis and natural language inference. CoRR, abs/2002.04815. ↗
  51. Naftali Tishby, Fernando C. Pereira, and William Bialek. 2000. The information bottleneck method. ↗
  52. Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, Aurelien Rodriguez, Armand Joulin, Edouard Grave, and Guillaume Lample. 2023. Llama: Open and efficient foundation language models. ↗
  53. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. CoRR, abs/1706.03762. ↗
  54. Hulin Wang, Donglin Yang, Yaqi Xia, Zheng Zhang, Qi-gang Wang, Jianping Fan, Xiaobo Zhou, and Dazhao Cheng. 2024a. Raptor-t: A fused and memory-efficient sparse transformer for long and variable-length sequences. IEEE Transactions on Computers, 73(7):1852–1865. ↗
  55. Qiqi Wang, Ruofan Wang, Kaiqi Zhao, Robert Amor, Benjamin Liu, Jiamou Liu, Xianda Zheng, and Zijian Huang. 2024b. SKGSum: Structured knowledge-guided document summarization. In Findings of ACL 2024, pages 1857–1871. find
  56. Chaojun Xiao, Xueyu Hu, Zhiyuan Liu, Cunchao Tu, and Maosong Sun. 2021. Lawformer: A pre-trained language model for chinese legal long documents. AI Open, 2:79–84. ↗
  57. Nuo Xu, Pinghui Wang, Long Chen, Li Pan, Xiaoyan Wang, and Junzhou Zhao. 2020. Distinguish confusing law articles for legal judgment prediction. In ACL, pages 3086–3095.
  58. Nuo Xu, Pinghui Wang, Junzhou Zhao, Feiyang Sun, Lin Lan, Jing Tao, Li Pan, and Xiaohong Guan. 2024. Distinguish confusion in legal judgment prediction via revised relation knowledge. 43(1). ↗
  59. Qi Xu, Qian Liu, Hao Fei, Hang Yu, Shuhao Guan, and Xiao Wei. 2025. Clear: A framework enabling large language models to discern confusing legal paragraphs. In Findings of the Association for Computational Linguistics: EMNLP 2025, pages 8937– 8953. find
  60. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, Chujie Zheng, Dayiheng Liu, Fan Zhou, Fei Huang, Feng Hu, Hao Ge, Haoran Wei, Huan Lin, Jialong Tang, and 41 others. 2025. Qwen3 technical report. Preprint, arXiv:2505.09388. ↗
  61. Junjie Yang and Hai Zhao. 2019. Deepening hidden representations from pre-trained language models for natural language understanding. CoRR, abs/1911.01940. ↗
  62. Shuo Yang, Ying Sheng, Joseph E. Gonzalez, Ion Stoica, and Lianmin Zheng. 2024. Post-training sparse attention with double sparsity. Preprint, arXiv:2408.07092. ↗
  63. Zichao Yang, Diyi Yang, Chris Dyer, Xiaodong He, Alex Smola, and Eduard H. Hovy. 2016. Hierarchical attention networks for document classification. In NAACL.
  64. Jingyang Yuan, Huazuo Gao, Damai Dai, Junyu Luo, Liang Zhao, Zhengyan Zhang, Zhenda Xie, Yuxing Wei, Lean Wang, Zhiping Xiao, Yuqing Wang, Chong Ruan, Ming Zhang, Wenfeng Liang, and Wangding Zeng. 2025. Native sparse attention: Hardware-aligned and natively trainable sparse attention. In ACL 2025, pages 23078–23097. ↗
  65. Linan Yue, Qi Liu, Binbin Jin, Han Wu, Kai Zhang, Yanqing An, Mingyue Cheng, Biao Yin, and Dayong Wu. 2021. Neurjudge: A circumstance-aware neural framework for legal judgment prediction. In SIGIR, page 973–982. find
  66. Manzil Zaheer, Guru Guruganesh, Kumar Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, and Amr Ahmed. 2020. Big bird: Transformers for longer sequences. In NeurIPS, volume 33, pages 17283–17297. ↗
  67. Qingru Zhang, Dhananjay Ram, Cole Hawkins, Sheng Zha, and Tuo Zhao. Efficient long-range transformers: You need to attend more, but not necessarily at every layer. In Findings of EMNLP 2023, pages 2775–2786. find
  68. Xingxing Zhang, Furu Wei, and Ming Zhou. 2019. HI-BERT: Document level pre-training of hierarchical bidirectional transformers for document summarization. In ACL 2019, pages 5059–5069. find
  69. Yu Zhong, Bo Shen, and Tao Wang. 2024. Tgin: Document-level event extraction with two-phase graph inference network. Neural Netw., 176:106343. Epub 2024 Apr 25. PMID: 38701598. ↗
  70. Yilun Zhou, Marco Túlio Ribeiro, and Julie Shah. 2022. Exsum: From local explanations to model understanding. In NAACL-HLT, pages 5359–5378. find
  71. Zhi Zhou, Kun-Yang Yu, Shi-Yu Tian, Xiao-Wen Yang, Jiang-Xin Shi, Pengxiao Song, Yi-Xuan Jin, Lan-Zhe Guo, and Yu-Feng Li. 2025. Lawgpt: Knowledge-guided data generation and its application to legal llm. Preprint, arXiv:2502.06572. ↗
  72. Liu Zhuang, Lin Wayne, Shi Ya, and Zhao Jun. 2021. A robustly optimized BERT pre-training approach with post-training. In CCL 2021, pages 1218–1227.
  73. Octavia-Maria S , ulea, Marcos Zampieri, Shervin Malmasi, Mihaela Vela, Liviu Dinu, and Josef Genabith. 2017. Exploring the use of text classification in the legal domain. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/lou-2026-less-is-more-knowledge-aware/