Research map / Large language models / AI and society
The Cambridge Law Corpus: A Dataset for Legal AI Research
Andreas Östling, Holli Sargeant, Huiyuan Xie, Ludwig Konrad Bull, Alexander Terenin, Leif Jonsson, Måns Magnusson, Felix Steffek
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| two-step RoBERTa pipeline instead of end-to-end token classification | RoBERTa · CLC case outcome extraction | F1: 0.185 -> 0.012 | Hurt |
| two-step RoBERTa pipeline instead of end-to-end token classification | RoBERTa · CLC case outcome extraction | accuracy: 0.997 -> 0.739 | Hurt |
| use GPT-4 instead of GPT-3.5-turbo (zero-shot) | zero-shot GPT · CLC case outcome extraction | ROUGE-L: 0.300 -> 0.517 | Improved |
| use GPT-4 instead of GPT-3.5-turbo (zero-shot) | zero-shot GPT · CLC case outcome extraction | WER: 4.281 -> 3.396 | Improved |
| use GPT-4 instead of GPT-3.5-turbo (zero-shot) | zero-shot GPT · CLC case outcome extraction | BLEU: 0.217 -> 0.282 | Improved |
| zero-shot GPT-4 instead of fine-tuned end-to-end RoBERTa | case outcome extractor · CLC case outcome extraction | BERTScore: 0.207 -> 0.840 | Improved |
| slice train/val/test sequentially by court instead of random shuffle (BERT) (ablation) | sentence-level classifier · CLC outcome sentence classification | extraction accuracy (%): 82.3 -> 50.8 | Hurt |
| slice train/val/test sequentially by court instead of random shuffle (T5) (ablation) | sentence-level classifier · CLC outcome sentence classification | extraction accuracy (%): 82.3 -> 53.9 | Hurt |
| T5 instead of BERT as sentence classifier under court-sliced split (ablation) | sentence-level classifier · CLC outcome sentence classification | extraction accuracy (%): 50.8 -> 53.9 | Improved |
Closest papers by meaning
- Lawma: The Power of Specialization for Legal Annotation (ICLR 2025 · similarity 0.89)
- Freeing the Law with LOCUS: A Local Ordinance Corpus for the United States (NeurIPS 2026 · similarity 0.87)
- LexEval: A Comprehensive Chinese Legal Benchmark for Evaluating Large Language Models (NeurIPS 2024 · similarity 0.87)
- SaulLM-54B & SaulLM-141B: Scaling Up Domain Adaptation for the Legal Domain (NeurIPS 2024 · similarity 0.86)
- LexChain: Modeling Legal Reasoning Chains for Chinese Tort Case Analysis (AAAI 2026 · similarity 0.84)
- Interpretable Long-Form Legal Question Answering with Retrieval-Augmented Large Language Models (AAAI 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
