In plain words: They kept the facts fixed, swapped the cited law for an unrelated one, and watched the verdict. Verdicts changed only 0-21.7% on one task despite models naming the right law 66.7-100% of the time, so citing a law doesn't prove the verdict rests on it.
Abstract · Cited but Not Consulted: A Counterfactual Audit of Legal Chain-of-Thought Faithfulness
Large language models increasingly justify legal decisions by naming the statute or precedent behind a verdict, treated as evidence that the decision follows from it. We test this directly: holding case facts fixed, we substitute the named legal authority for an unrelated one and decode a model's evolving verdict from its hidden states. Across seven open-weight models (8B-70B) and four benchmarks spanning judicial and contractual reasoning, when explicitly required to justify a verdict by naming the governing authority, models name the correct one in 66.7%-100% of generations, while the verdict changing when the authority changes is far less consistent: 0.0%-21.7% on CaseHOLD, 30.0%-76.7% on ECHR and SCOTUS, and 43.3%-50.0% on ContractNLI. Neither scale nor a purpose-built legal-reasoning model (a best-effort LoRA reproduction; Section 6) closes this gap. A red-teaming evaluation on five core models finds compliance with an adversarial instruction hidden in the case facts (73.3%-96.4%) exceeds verdict-swap sensitivity by a wide margin, holding without exception across model rankings. Naming a legal authority is thus a poor proxy for a verdict's dependence on it, while the same verdict remains separately vulnerable to adversarial manipulation. Both findings replicate across checks ruling out prompt-wording noise and confounded sampling, and bear directly on the use of generated legal explanations as compliance or audit artefacts.
Saisab Sadhu, Shreeyans Arora, Pratinav Seth
arXiv:2610.12361 · cs.AI, cs.CL, cs.CY · submitted Oct 8, 2026
abstract · pdf · html