Large language models increasingly justify legal decisions by naming the statute or precedent behind a verdict, treated as evidence 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 regardless of cross-model ranking. Naming a legal authority is thus a poor proxy for a verdict’s dependence on it, and 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 using generated legal explanations as compliance or audit artefacts.

Large language model compliance systems are deployed on the assumption that a verdict depends on the regulatory rule it is given. We test this directly across five models and 20 regulatory and platform-policy domains: delete, swap, or negate the governing rule while holding the case fixed, and check whether the verdict changes (OCS) or the model’s internal representation of compliance shifts at all (ICS-delta). Neither moves much: models’ verdicts are often invariant to substantial perturbations of the supplied rule, and the guard model, evaluated here under a custom-rule adaptation of its native taxonomy, is the least rule-sensitive and least accurate of the five, barely above chance (51%, versus 90–92% for general-purpose models). This reflects easy cases more than blanket neglect: on cases where deleting the rule changes a previously correct model prediction, models do track it closely. Neither better prompting nor direct intervention on the model’s internal representations closes this gap. Accuracy alone does not establish that a compliance verdict is grounded in the supplied rule.

Methods for repairing safety alignment in fine-tuned Large Language Models (LLMs) are evolving quickly, but the literature is fragmented across the repair pipeline: each method is typically tested on its own drifted checkpoint, with its own repair implementation, and judged against its own set of baselines. This makes methods hard to compare or select, and calls for a framework that standardizes drift-induction, repair, and evaluation while remaining flexible enough to support new paradigms and architectures. To address this gap, we introduce SAFETUNE, a source-available library that unifies drift-induction, repair, and evaluation behind a standardized pipeline. SAFETUNE hosts four repair paradigms: weight patching, safety-constrained retraining, gradient-based unlearning, and inference-time steering, with implementations audited and debugged against their original papers. A documented intervention guide suggests a starting repair for a checkpoint’s safety-capability profile, with the full pipeline runnable in four lines of code.

The efficacy of state-of-the-art abstractive summarization models is severely constrained by the extreme document lengths of legal judgments, which consistently surpass their fixed input capacities. The prevailing method, naive sequential chunking, is a discourse-agnostic process that induces context fragmentation and degrades summary coherence. This paper introduces Structure-Aware Chunking (SAC), a rhetorically-informed pre-processing pipeline that leverages the intrinsic logical structure of legal documents. We partition judgments into their constituent rhetorical strata—Facts, Arguments & Analysis, and Conclusion—prior to the summarization pass. We present and evaluate two SAC instantiations: a computationally efficient heuristic-based segmenter and a semantically robust LLM-driven approach. Empirical validation on the JUST-NLP 2025 L-SUMM shared task dataset reveals a nuanced trade-off: while our methods improve local, n-gram based metrics (ROUGE-2), they struggle to maintain global coherence (ROUGE-L). We identify this “coherence gap” as a critical challenge in chunk-based summarization and show that advanced LLM-based segmentation begins to bridge it.