Legal QA benchmarks have predominantly focused on case law, overlooking the unique challenges of statute-centric regulatory reasoning. In statutory domains, relevant evidence is distributed across hierarchically linked documents, creating a statutory retrieval gap where conventional retrievers fail and models often hallucinate under incomplete context. We introduce SearchFireSafety, a structure- and safety-aware benchmark for statute-centric legal QA. Instantiated on fire-safety regulations as a representative case, the benchmark evaluates whether models can retrieve hierarchically fragmented evidence and safely abstain when statutory context is insufficient. SearchFireSafety adopts a dual-track evaluation framework combining real-world questions that require citation-aware retrieval and synthetic partial-context scenarios that stress-test hallucination and refusal behavior. Experiments across multiple large language models show that graph-guided retrieval substantially improves performance, but also reveal a critical safety trade-off: domain-adapted models are more likely to hallucinate when key statutory evidence is missing. Our findings highlight the need for benchmarks that jointly evaluate hierarchical retrieval and model safety in statute-centric regulatory settings.

Improving the reasoning abilities of large language models (LLMs) has largely relied on iterative self-training with model-generated data. While effective at boosting accuracy, existing approaches primarily reinforce successful reasoning paths, incurring a substantial calibration cost: models become overconfident and lose the ability to represent uncertainty. This failure has been characterized as a form of model collapse in alignment, where predictive distributions degenerate toward low-variance point estimates.We address this issue by reframing open-ended reasoning training as an epistemic learning problem, in which models must learn not only how to reason, but also when their reasoning should be trusted. We propose epistemically-calibrated reasoning (EpiCaR) as a training objective that jointly optimizes reasoning performance and calibration, and instantiate it within an iterative supervised fine-tuning framework using explicitly extracted meta-cognitive self-evaluation signals. Experiments on Llama-3 and Qwen-3 families demonstrate that our approach achieves Pareto-superiority over standard baselines in both accuracy and calibration, particularly in models with sufficient reasoning capacity (e.g., 3B+). This framework generalizes effectively to OOD mathematical reasoning (GSM8K) and code generation (MBPP). Ultimately, our approach enables a 3× reduction in the overall inference compute budget, matching the K=30 majority-vote performance of STaR with only K=10 confidence-weighted samples, entirely without the multi-model overhead of external verifiers.

In recent years, significant advancements in conversational question and answering (CQA) have been driven by the exponential growth of large language models and the integration of retrieval mechanisms that leverage external knowledge to generate accurate and contextually relevant responses. Consequently, the fields of conversational search and retrieval-augmented generation (RAG) have obtained substantial attention for their capacity to address two key challenges: query rewriting within conversational histories for better retrieval performance and generating responses by employing retrieved knowledge. However, both fields are often independently studied, and comprehensive study on entire systems remains underexplored. In this work, we present a novel retrieval-augmented conversation (RAC) dataset and develop a baseline system comprising query rewriting, retrieval, reranking, and response generation stages. Experimental results demonstrate the competitiveness of the system and extensive analyses are conducted to apprehend the impact of retrieval results to response generation.