[KongchangAI]
PaperClassifying Interpretive Canons at the Sentence Level

arXiv:2609.26945v1

A research paper published on arXiv that proposes a sentence-level benchmark for classifying legal interpretation canons based on decisions of the German Federal Constitutional Court, and evaluates the ability of large language models to identify legal interpretation methods.

Timeline (last 90 days)

Sep 24

An arXiv study (arXiv:2609.26945v1) proposed a sentence-level evaluation benchmark to test the ability of large language models to identify legal interpretation canons.

Unverified70%

All Facts (1)

Source Articles