ArabLegalEval
An evaluation benchmark for the Saudi Arabic legal domain, used to assess the performance of large language models on legal question-answering tasks. It supports two evaluation protocols: with context and without context.
Timeline (last 90 days)
After removing context, the accuracy gap of the base models on the legal benchmark can reach up to 27 percentage points.
The study examined the performance of three stages—domain adaptation, quantization, and knowledge distillation—on three models: FANAR-9B, JAIS-8B, and Qwen2.5-7B.
All Facts (2)
After removing context, the accuracy gap of the base models on the legal benchmark can reach up to 27 percentage points.
50%UnverifiedThe study examined the performance of three stages—domain adaptation, quantization, and knowledge distillation—on three models: FANAR-9B, JAIS-8B, and Qwen2.5-7B.
50%