anonymous commited on
Commit
916a451
·
1 Parent(s): 2e511b5

Add card metadata to the bundle README

Browse files
Files changed (1) hide show
  1. README.md +26 -13
README.md CHANGED
@@ -1,14 +1,28 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  # LEGEX — Reproduction Bundle (release/icml2026)
2
 
3
- LEGEX is an expert-coded benchmark for civil-judgment review-table
4
  extraction: judgments from the highest civil courts of 19 jurisdictions,
5
  annotated by legal experts on a shared schema (monetary outcomes, cost
6
  allocation, party structure, industry classification), against which we
7
  evaluate two commercial review-table systems (Harvey, Legora) and two
8
  schema-constrained LLM pipelines (Gemini, ChatGPT). This bundle is the
9
- versioned code + data payload behind the paper LEGEX: An Eight-Jurisdiction
10
- Benchmark for Legal Review-Table Extraction (ICML 2026 Workshop on AI for
11
- Law). It contains an eight-jurisdiction core benchmark (≥100 judgments each,
12
  double-annotated on a 28–30-case overlap) and eleven preview jurisdictions.
13
 
14
  The release consists of three Hugging Face repositories, pinned to the
@@ -101,18 +115,17 @@ bash scripts/reproduce_paper.sh
101
  git status data/analysis
102
  ```
103
 
104
- ## Unpublished ata
105
 
106
  The expert-annotated XLSX workbooks (`data/<cc>/Goldenset_*_final*.xlsx`)
107
- behind the goldensets arenot published: Microsoft Office embeds author
108
  metadata, and the re-annotation assignment structure could identify
109
- individual annotators. The gold labels are published in full as
110
- `goldensets/data/<cc>/goldenset_<cc>.jsonl` with primary + anonymized secondary
111
- annotator rows, and every analysis in this bundle runs from those published
112
- files, the workbooks are not needed for reproduction.
113
 
114
- ## Alternative Annotator Test (AAT).
115
- The AAT (Calderon, Reichart & Dror,
116
  ACL 2025, [arXiv:2501.10970](https://arxiv.org/abs/2501.10970)) is run with
117
  the authors' original implementation via the adapter
118
  `scripts/alt_test_reference.py`, which extracts `alt_test()` from the
@@ -137,7 +150,7 @@ The scrapers that built the corpus are included (`legex/scrapers/`, one file
137
  per jurisdiction; see `docs/Jurisdictions.md` for the per-country catalogue).
138
 
139
  ```bash
140
- cp .env.template .env
141
  uv run legex-run
142
  ```
143
 
 
1
+ ---
2
+ license: mit
3
+ pretty_name: "LEGEX Reproduction Bundle: Code, Pipeline, and Analysis Outputs"
4
+ language:
5
+ - en
6
+ tags:
7
+ - legal
8
+ - benchmark
9
+ - llm-evaluation
10
+ - information-extraction
11
+ - reproducibility
12
+ viewer: false
13
+ ---
14
+
15
  # LEGEX — Reproduction Bundle (release/icml2026)
16
 
17
+ **LEGEX** is an expert-coded benchmark for civil-judgment review-table
18
  extraction: judgments from the highest civil courts of 19 jurisdictions,
19
  annotated by legal experts on a shared schema (monetary outcomes, cost
20
  allocation, party structure, industry classification), against which we
21
  evaluate two commercial review-table systems (Harvey, Legora) and two
22
  schema-constrained LLM pipelines (Gemini, ChatGPT). This bundle is the
23
+ versioned code + data payload behind the paper *LEGEX: An Eight-Jurisdiction
24
+ Benchmark for Legal Review-Table Extraction* (ICML 2026 Workshop on AI for
25
+ Law); it contains an eight-jurisdiction core benchmark (≥100 judgments each,
26
  double-annotated on a 28–30-case overlap) and eleven preview jurisdictions.
27
 
28
  The release consists of three Hugging Face repositories, pinned to the
 
115
  git status data/analysis
116
  ```
117
 
118
+ ## Unpublished raw data — scope
119
 
120
  The expert-annotated XLSX workbooks (`data/<cc>/Goldenset_*_final*.xlsx`)
121
+ behind the goldensets are **not published**: Microsoft Office embeds author
122
  metadata, and the re-annotation assignment structure could identify
123
+ individual annotators. The gold *labels* are published in full as
124
+ `goldensets/data/<cc>/goldenset_<cc>.jsonl` (primary + anonymized secondary
125
+ annotator rows), and every analysis in this bundle runs from those published
126
+ files the workbooks are not needed for reproduction.
127
 
128
+ **Alternative Annotator Test (AAT).** The AAT (Calderon, Reichart & Dror,
 
129
  ACL 2025, [arXiv:2501.10970](https://arxiv.org/abs/2501.10970)) is run with
130
  the authors' original implementation via the adapter
131
  `scripts/alt_test_reference.py`, which extracts `alt_test()` from the
 
150
  per jurisdiction; see `docs/Jurisdictions.md` for the per-country catalogue).
151
 
152
  ```bash
153
+ cp .env.template .env # then fill in what you need
154
  uv run legex-run
155
  ```
156