Large Language Models (LLMs) often exhibit "Attention Sink" (AS) and the accompanying "Massive Activations" (MAs) at the initial position of a sequence. These phenomena frequently co-occur, and MAs can pose challenges for low-bit quantization. In this study, we analyze the factors underlying AS and MAs that emerge at the initial position regardless of the token occupying it. Our experiments suggest that self-concentration of attention, resulting from the causal mask, and the subsequent Value-non-mixing in attention outputs contribute to AS and MAs. These findings provide new empirical evidence on the internal dynamics of LLMs, offering insights that may inform future quantization strategies and advance our understanding of the internal mechanisms of attention layers.

We evaluate the alignment of large language models (LLMs) and large vision-language models (LVLMs) with human perception, focusing on the Japanese concept of shitsukan, which reflects the sensory experience of perceiving objects. We created a dataset of shitsukan terms elicited from individuals in response to object images. With it, we designed benchmark tasks for three dimensions of understanding shitsukan: (1) accurate perception in object images, (2) commonsense knowledge of typical shitsukan terms for objects, and (3) distinction of valid shitsukan terms. Models demonstrated mixed accuracy across benchmark tasks, with limited overlap between model- and human-generated terms. However, manual evaluations revealed that the model-generated terms were still natural to humans. This work identifies gaps in culture-specific understanding and contributes to aligning models with human sensory perception. We publicly release the dataset to encourage further research in this area.

Human children acquire language from a substantially smaller amount of linguistic input than that typically required for training large language models (LLMs). This gap motivates the search for more efficient pre-training methods. Inspired by child development, curriculum learning, which progresses from simple to complex data, has been widely adopted. In this study, we propose a pre-training framework that mirrors child language acquisition, advancing step by step from words to sentences while retaining prior knowledge. We investigate whether this improves retention and efficiency under limited resources. Our approach is implemented through four components: (i) a curriculum-aligned dataset, (ii) a batch-wise convergence loop, (iii) a distance-controlled loss to mitigate forgetting, and (iv) a constraint-controlled optimizer for stability. Experiments on the BabyLM benchmark show that the proposed method performs slightly below the official baselines in overall accuracy, with larger gaps on grammar-oriented evaluations such as BLiMP. Nonetheless, it yields small but consistent gains on morphology- and discourse-related tasks (e.g., WUG-ADJ, Entity Tracking), suggesting that the approach affects different linguistic aspects unevenly under limited data conditions.

With the remarkable development of large language models (LLMs), ensuring the factuality of output has become a challenge.However, having all the contents of the response with given knowledge or facts is not necessarily a good thing in dialogues.This study aimed to achieve both attractiveness and factuality in a dialogue response for which a task was set to predict sentences that do not require factual correctness judgment such as agreeing, or personal opinions/feelings.We created a dataset, dialogue dataset annotated with fact-check-needed label (DDFC), for this task via crowdsourcing, and classification tasks were performed on several models using this dataset.The model with the highest classification accuracy could yield about 88% accurate classification results.