A legal reasoning dataset prepared for the Reasoning Datasets Competition by Bespoke Labs, Hugging Face, and Together.ai. It contains multilevel human and model-annotated explanations, focusing on international acts and treaties to support jurisdiction-agnostic evaluation. The dataset was last updated on May 1, -2025.
Use Cases
- Training legal reasoning models based on the provided multilevel explanations.
- Benchmarking AI systems on multilingual legal interpretation tasks based on the focus on international acts.
- Developing explanation generation models for legal text based on the human and model-annotated structure.
Strengths
- Includes multilevel explanations, combining both human and model annotations.
- Focuses on international acts and treaties, suggesting a jurisdiction-agnostic scope.
- Created for a structured competition involving Bespoke Labs, Hugging Face, and Together.ai.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
- Source
- ciol-research on Hugging Face.
- Collection Method
- Prepared as part of the Reasoning Datasets Competition.
- Time Range
- null
- Freshness
- Last updated 2025-05-01 19:03:46; freshness should be verified.
- Geography
- International, based on the focus on international acts and treaties.