Newest papers

Commit or Abstain? System One Decision Models, LLMs, and Missing Legal Authority

M. Mikail Demir · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.29961 · cited by 0 works (OpenAlex) · open-access PDF

Indonesian

Mengingat jumlah besar klaim yang harus diperiksa, AI hukum yang dapat diperiksa membutuhkan model yang dapat diperluas, efisien biaya yang memenuhi kebutuhan praktis hukum. Model keputusan System One dibangun pada model bahasa besar tetapi, alih-alih menghasilkan teks, mengembalikan probabilitas terkalibrasi untuk masing-masing opsi. Ini cocok untuk verifikasi otoritas hukum, di mana pertanyaan ditutup: holding tertentu, atau tidak ada dari mereka. Kami menguji apakah model tersebut menahan diri ketika otoritas yang benar hilang. Kami memperluas CaseHOLD dengan kondisi abstensi di mana holding yang benar diganti di 1.670 dari 5.314 item, dan membandingkan empat model keputusan dengan enam LLM generatif. Ketika tidak ada pilihan yang benar, LLM generatif masih memilih satu di 38-86% item; model keputusan yang paling hati-hati melakukannya di 30%. Pada CaseHOLD standar, model keputusan menilai

Machine translation (Meta M2M100); the English original follows.

Abstract

Given the enormous number of claims to be checked, verifiable legal AI requires scalable, cost-effective models that answer the practical needs of law. System One decision models are built on large language models but, instead of generating text, return a calibrated probability for each option. This suits the verification of legal authority, where the question is closed: a given holding, or none of them. We test whether such models abstain when the correct authority is missing. We extend CaseHOLD with an abstention condition in which the correct holding is replaced in 1,670 of 5,314 items, and compare four decision models with six generative LLMs. When no option is correct, generative LLMs still pick one in 38-86% of items; the most cautious decision model does so in 30%. On standard CaseHOLD, a decision model matches the best generative LLM (88.1% vs. 88.2%) at 1/55 of the cost. Accepting that LLM's answer only when a decision model agrees halves its wrong citations, and distilling a decision model's probabilities into Qwen3-8B, without labels, yields a well-calibrated local model. Legal AI should be judged by its wrong citations, not by accuracy alone.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Commit, Computer science, Generative grammar, Decision model, Verifiable secret sharing, Decision analysis

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange