LexRevise-MT v1.1: a benchmark for Legal Conclusion Revision under supervening evidence in multi-turn consultations

LexRevise-MT v1.1 is a benchmark for Legal Conclusion Revision (LCR): the per-claim, per-turn classification of whether a legal conclusion asserted in a multi-turn consultation should be maintained, revised, withdrawn, or reinstated as supervening evidence arrives. The release contains 250 four-turn scenarios and 4,609 claim-turn labels over the operators NEW, PERSIST, REVISE and RETRACT in Spanish succession law, with scenario-level and arc-level splits, five supervening-event types and 47 traceable legal sources. It includes the corpus, the deterministic-baseline and inter-annotator-agreement code (Python, standard library only), the released results, the dataset card and codebook, and a SHA-256 manifest. Reliability on 650 jurist-adjudicated items: three-class Fleiss κ = 0.803; four-class κ = 0.711. See README.md for reproduction steps