Where Do Legal Factors Live in LLMs?
In AI & Law, HYPO dimensions and CATO factors represent these patterns symbolically, organizing expert knowledge distilled from case narratives as explicit, structured, interpretable, and auditable legal concepts and relations that support case comparison, legal argument, and precedent-based reasoning.
ProofPaper ↗
Key points
- In case-based legal reasoning, legal professionals identify legally significant fact patterns in a new dispute and compare them with those of decided cases.
- Large language models (LLMs) are increasingly used for legal tasks, yet their internal organization of legal knowledge remains difficult to trace.
- We investigate whether LLMs form internal representations corresponding to all 26 CATO factors.
- By locating these readouts, our study connects legal language and the symbolic abstractions of HYPO and CATO to learned internal representations.
Sources (1)
- [1]Where Do Legal Factors Live in LLMs?arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 20, 06:24 PM
In AI & Law, HYPO dimensions and CATO factors represent these patterns symbolically, organizing expert knowledge distilled from case narratives as explicit, structured, interpretable, and auditable legal concepts and relations that support case comparison, legal argument, and precedent-based reasoning.
In case-based legal reasoning, legal professionals identify legally significant fact patterns in a new dispute and compare them with those of decided cases.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 19, 2026Shaped SAEs: Expanding the Toolkit for Mechanistic Interpretability
- Oct 19, 2026Magnolia: A Persistent-Memory Agent Framework for Computational Chemistry
- Oct 19, 2026The Culture Within: Towards Mechanistic Understanding of Cultural Reasoning in LLMs
- Oct 19, 2026Safety Mid-Training: Internalizing LLM Safety as a Foundational Capability
- Oct 19, 2026Introducing Clairic, a new System One class of extraction models
- Oct 19, 2026From Black Box to Controller: Inference-Time Steering of LLM via Feature Amplification