Probabilistic decoding in Large Language Models (LLMs) often yields inconsistent outputs, particularly on complex or long-form questions. Self-Consistency (SC) mitigates this for short-form QA by majority voting over exact strings, whereas Universal Self-Consistency (USC) and Weighted Unigram Consistency Score (WUCS) extend to long-form responses but lose accuracy on short-form benchmarks. We introduce Latent Self-Consistency (LSC), which selects the most semantically consistent response using learnable token embeddings. LSC's lightweight forward processing of summary tokens only introduces negligible runtime overhead (at most 0.9%) on top of standard decoding of the base LLM, and requires no changes to the model architecture. Across 6 short-form and 5 long-form reasoning benchmarks (e.g., MATH, MMLU, TruthfulQA), LSC surpasses SC, USC, and WUCS on both short-form and long-form on average performance, while adding negligible computational overhead on vanilla inference. These results position LSC as a reliable consistency-selection method that works effectively across various answer formats. Additionally, LSC provides well-calibrated confidence estimates, maintaining low expected calibration error across both answer formats.

This research proposes a novel method of measuring the dynamics of semantic relatedness. Research on semantic relatedness has a long history in the fields of computational linguistics, psychology, computer science, as well as information systems. Computing semantic relatedness has played a critical role in various situations, such as data integration and keyword recommendation. Many researchers have tried to propose more sophisticated techniques to measure semantic relatedness. However, little research has considered the change of semantic relatedness with the flow of time and occurrence of events. The authors' proposed method is validated by actual corpus data collected from a particular context over a specific period of time. They test the feasibility of our proposed method by constructing semantic networks by using the corpus collected during a different period of time. The experiment results show that our method can detect and manage the changes in semantic relatedness between concepts. Based on the results, the authors discuss the need for a dynamic semantic relatedness paradigm.