Research map / Large language models / Language, safety and interpretability
Multi-LexSum: Real-world Summaries of Civil Rights Lawsuits at Multiple Granularities
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| use long-input models (LED, PRIMERA) instead of BART/PEGASUS | abstractive summarizers · Multi-LexSum D->L | ROUGE-2 F1: 23.78 -> 27.32 | Helped |
| increase LED input length from 4096 to 16384 (ablation) | LED · Multi-LexSum D->T | ROUGE-2 F1: 8.92 -> 9.84 | Helped |
| abstractive models instead of extractive baselines | summarization · Multi-LexSum D->L | ROUGE-2 F1: 10.65 -> 20.01 | Helped |
| summarize from a gold long summary instead of source documents | BART/PEGASUS · Multi-LexSum L->S | ROUGE-2 F1: 19.98 -> 37.02 | Helped |
| replace gold summary input with BART-predicted summary in a pipeline (ablation) | PRIMERA {S,D}->T · Multi-LexSum tiny | ROUGE-2 F1: 16.50 -> 6.79 | Hurt |
| multitask training on L, S and T with prefixes instead of single-task (ablation) | BART · Multi-LexSum D->L | ROUGE-2 F1: 20.01 -> 23.24 | Helped |
| multitask training on L, S and T instead of single-task (ablation) | BART · Multi-LexSum D->T | ROUGE-2 F1: 7.09 -> 8.92 | Helped |
| multitask training instead of single-task (ablation) | BART · Multi-LexSum D->S | ROUGE-2 F1: 19.91 -> 20.14 | No clear effect |
| two-task multitask training for L->{S,T} (ablation) | BART · Multi-LexSum L->S | ROUGE-2 F1: 37.02 -> 36.11 | Hurt |
Closest papers by meaning
- Lawma: The Power of Specialization for Legal Annotation (ICLR 2025 · similarity 0.81)
- Freeing the Law with LOCUS: A Local Ordinance Corpus for the United States (NeurIPS 2026 · similarity 0.81)
- LexEval: A Comprehensive Chinese Legal Benchmark for Evaluating Large Language Models (NeurIPS 2024 · similarity 0.81)
- BooookScore: A systematic exploration of book-length summarization in the era of LLMs (ICLR 2024 · similarity 0.80)
- EHRNoteQA: An LLM Benchmark for Real-World Clinical Practice Using Discharge Summaries (NeurIPS 2024 · similarity 0.80)
- QuerySum: A Multi-Document Query-Focused Summarization Dataset Augmented with Similar Query Clusters (AAAI 2024 · similarity 0.80)
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.