The field of large language models (LLMs) is moving towards increased transparency and accountability, with a focus on developing methods to audit and improve their reasoning processes. Recent research has introduced novel approaches to analyzing and evaluating the semantic structures of complex reasoning traces, enabling the identification of distinct reasoning patterns and the detection of probable reasoning errors. Furthermore, innovations in hallucination detection have led to the development of new frameworks that leverage perturbation sensitivity and specialized model divergence to separate truthful from hallucinated responses. Additionally, there is a growing interest in exploring the limitations of LLMs in translating logical reasoning across lexically diversified contexts and in developing fine-grained evaluation frameworks to assess logical reasoning capabilities. Notable papers in this area include: ReasoningFlow, which introduces a unified schema for analyzing the semantic structures of complex reasoning traces. Shaking to Reveal, which proposes a perturbation-based method for detecting LLM hallucinations. Ask a Local, which exploits the intuition that specialized models exhibit greater surprise when encountering domain-specific inaccuracies to detect hallucinations.