This study explores the implementation of human-like memory behavior in language agents by integrating the ACT-R cognitive architecture with large language models (LLMs). We designed a dialogue agent capable of dynamically retrieving and forgetting memories based on context, time, and usage frequency. The proposed system utilizes a vector-based activation mechanism, incorporating temporal decay, semantic similarity, and probabilistic noise to mimic natural memory dynamics. Simulation experiments confirmed the model’s ability to reproduce memory reinforcement through repeated topics, as well as stochastic variability in memory retrieval, reflecting human memory behavior. We also identified optimal parameters for balancing contextual sensitivity and memory stability. This work contributes to the development of more human-compatible AI dialogue systems by modeling memory not as mere storage, but as a strategic, context-sensitive process.

Picture-book storytelling fosters preschool children's cognitive and literacy development in Japanese childcare centers. However, training childcare workers remains challenging due to the need for emotional engagement and adaptive strategies. This study develops a VR-based storytelling training system that enables trainees to practice in an immersive environment with virtual child avatars. The system evaluates performance based on voice volume, reading speed, posture, gaze, and eye contact, providing real-time feedback. An expert evaluation confirmed the system's validity but identified areas for improvement. Adaptive scoring for voice and reading speed, tracking of page-turning techniques, and body orientation assessment should be incorporated. Additionally, realistic child reactions and adjusted gaze evaluation for children sitting on the floor are needed.

Sustaining user memory in digital environments such as virtual reality (VR) is a significant challenge. We show that temporary tempo modulations in background music (BGM) can selectively and naturally enhance memory in VR. In a user study (N = 20), decreasing the BGM tempo by approximately 21% significantly improved recall of events. These findings point to a new acoustic design approach that adapts scene by scene to narrative pacing and importance while maintaining a natural user experience.

Since office work has become large-scale and diversified in companies or organizations, work engagement and efficiency have been always an important index of a team's or group's evaluation because it is directly connected to their outcomes. In order to identify the group work context, we first need to recognize for what and how long the individual members are spending their time at their desks, but without privacy concerns and underestimation of their actual work. In this paper, we propose and evaluate the base system of personal desk activity recognition by using a low-cost compact WiFi node and its WiFi channel state information (CSI), which can lead to a lightweight group work context identification system. As a result, we achieved 94.2% desk activity recognition accuracy using the on-desk receiver, in recognizing five different classes.

In this paper, we propose ToonMeet, a hybrid frame-work for high-resolution and style-controllable online meeting toonification that ensures real-time operation speed. ToonMeet applies video frame interpolation to traditional portrait toonification pipelines, allowing for the synthesis of intermediate frames between adjacent toonified keyframes, significantly accelerating the overall process and saving computational resources. However, this approach brings a new problem, where prevailing flow-based video frame interpolation methods tend to cause more ghost and blur artifacts in toonified scenes compared to non-toonified scenes, especially when fast-moving objects exist. We study this previously undiscussed problem and explore its causes. To address this, we introduce a new dataset called TM3B (Toonified Multi-modal Meeting Behaviors), offering high-resolution and cross-platform multi-modal stylized meeting data of Japanese youth in various scenarios. Then, we fine-tune ToonMeet on these tailored data and the resulting model presents improved optical flow estimation ability on toonified videos. Extensive experiments demonstrate that ToonMeet can achieve great spatiotemporal performance and perform high-quality toonification of online meetings with real-time operation speed.

Confluence of extreme-scale quantum dynamics simulations (i.e. [email protected]) and cutting-edge x-ray free-electron laser experiments are revolutionizing materials science. An archetypal example is the exciting concept of using picosecond light pulses to control emergent material properties on demand in atomically-thin layered materials. This paper describes efforts to scale our quantum molecular dynamics engine toward the United States' first exaflop/s computer, under an Aurora Early Science Program project named "Metascalable layered material genome". Key algorithmic and computing techniques incorporated are: (1) globally-scalable and locally-fast solvers within a linear-scaling divide-conquer-recombine algorithmic framework; (2) algebraic 'BLASification' of computational kernels; and (3) data alignment and loop restructuring, along with register and cache blocking, for enhanced vectorization and efficient memory access. The resulting weak-scaling parallel efficiency was 0.93 on 131,072 Intel Xeon Phi cores for a 56.6 million atom (or 169 million valence-electron) system, whereas the various code transformations achieved 5-fold speedup. The optimized simulation engine allowed us for the first time to establish a significant effect of substrate on the dynamics of layered material upon electronic excitation.

Augmented reality (AR) optical-see-through (OST) head-mounted displays (HMD) have developed to a point where browsing information on the go is possible. In this paper, we investigate the readability of text on an AR HMD while the user is walking. There are two common methods of displaying text on a HMD: anchoring the text on the screen coordinate system or the world coordinate system. We report on the results of two laboratory experiments comparing text readability when the text is displayed in these two coordinate systems, and while the participants walked on a treadmill. In the first experiment, the participants read letter strings comprising Sloane letters, whereas the second experiment used English words. In addition to evaluating the text readability and workload experienced by participants, we employed IMU sensors to compare the effects of the text display method on the participants' head movement and gait. In both experiments, the reading speed and head movement were significantly higher and mental workload significantly lower for the world coordinate system than for the screen coordinate system. These results suggest that text readability while walking is better on the world coordinate system, and displaying text with the screen coordinate system results in an unnatural gait owing to the user trying to keep their head still in an effort to stabilize the HMD screen.

The projection of invisible data on visible images can facilitate seamless interactive projection, since data embedded in regular images is unobtrusive to human viewers. However, the previous techniques sacrificed one of the following key goals: 1) calibration-free setup; 2) full-color projection; or 3) high contrast image. In this paper, we propose a Pixel-level Infrared Light Communication (PILC) projector that achieves all these requirements by adding an infrared light source to the full-color projector. To provide a proof of concept, we built a functional prototype, evaluated its performance, and presented a basic application.

Video records are sometimes used during meetings nowadays in order to compensate for the incompleteness of hand-written minutes. To use video effectively to record the minutes of a meeting, indices are considered necessary in addition to standard functions of movie players. Some systems can form indices by sensing human actions in the world, but such indices are not directly relevant to the contents of the discussions that occur in meetings. In this study, we propose an automated meeting index system based on people's discussions during daily meetings in the world. In our system, indices based on the contents of discussion can be automatically formulated in meetings if only paper, writing implements, and a camera are arranged in a certain manner. To this end, we develop trackable sticky notes, called "Tracky Notes," as indices. Tracky Notes can be tracked with visible markers through a camera, and is robust against occlusion. We then propose a meeting viewer where Tracky Notes are used as indices. Finally, we report the result of a user study conducted in a university classroom. The study revealed the appropriate region of visible markers and the robustness against occlusion of Tracky Notes. We also measured the effectiveness of our system comparing with standard functions of movie players.

Nodding has various communicative functions in humans, such as agreement, emphasis and turn-taking and can also create various positive impressions in communication by the person exhibiting the behavior. The ultimate aim of our research is to facilitate communication by implementing nodding behavior in robots. This study analyzed videos of human conversations in groups and focused on three aspects of people's nodding behavior as they listened to others: 1) Time period to complete a nodding cycle and each nods, 2) Time delay before initiating a nodding response, and 3) Number of continuous nods used at one time. We found that: 1) The mode time period to a nod was 0.27 s, with 96% of all nods occurring within 0.17~0.57 s. 2) The mode time delay before initiating a nodding response was 0.30 s, with 95% of all nods occurring within -0.78~1.4 s. 3) Fewer than six continuous nods were used 97% of the time - one nod, 55%, two nods, 24%, three nods, 12%, four nods, 3.0%, and five nods, 2.1%. Ultimately, the research findings serve as guidelines for implementing accurate human nodding behavior in robots.

Sensing the position and movement of the accelerator pedal in a vehicle is important for acceleration control and safety while driving. The accelerator pedal is controlled by the foot, but precise adjustment requires much training because the driver must rely on somatosensory cues, which provide limited feedback. In this study, we propose periodic tactile feedback for the accelerator pedal to provide an additional tactile cue. We conducted an experiment using a driving simulator to compare the lap time, the rate of off-track incidents and the subjective evaluation of controllability recorded in questionnaires. The experiment confirmed that the feedback makes the control of acceleration easier and facilitates safer driving.

With the aim of augmenting auditory sensation by tactile stimuli, we investigated cross-modal relationships between the two modalities, focusing on frequency. The results showed that frequency consonance between tactile and audio stimuli depends on the relationship between harmonics, in a manner similar to auditory waves, but with broader peaks.

Characters with body materials that are different from that of humans, such as metal robots or rubber people, frequently appear in movies and comics. While the abilities of their synthetic bodies can be easily observed from their actions, their somatic sensations are more difficult to appreciat e. Our aim in this work is to simulate the alteration of the material of the human body by means of vibrotactile feedback. The feedback represents the properties of the materials and is periodically applied to the elbow joint in synchrony with the elbow angle. This simulated sensation of having a different body material gives us the feeling of those characters. This technique can also be applied to improve maneuverability in the teleoperation of master-slave systems because it gives the operator a robot-like sensation.

Along public pathways, visual signs and audio cues are used by pedestrians to guide them into forming smoother pedestrian flows. However, often ignored or neglected, these signals require greater pedestrian attentiveness and appropriate conscious effort. To solve this problem, we have proposed the concept of "vection field". This is a field of optical flow that cues movement according to a pedestrian's motion. Visual stimulus within this optical flow leads pedestrians innately in specific directions without requiring direct interventions. We have implemented such a field by covering the ground with a lenticular lens screen; in this setup, neither power supply nor position tracking of pedestrians is necessary. An experimental result from our previous study shows that a vection field can direct pedestrians to one side. However, the quality of the optical flow such as image clarity and smoothness of motion was unsatisfactory in that it could cause a reduction in leading inducement. In this paper, we describe in detail a new display method involving a lenticular lens screen that yields an improvement in the quality of the vection field and ultimately pedestrian optical flow. Experiments showed improvements over previous attempts.

We propose a tactile enhancement structure inspired by hair follicle receptors. Unlike other receptors, part of the hair follicle receptor is exposed to the outside. Recent research has shown that skin hair contributes to perception of minute forces that cannot be perceived with glabrous skin. We considered how the skin perceives these minute forces. Our tactile enhancement device mimics the structure of hair follicle receptors. A matrix structure simulating artificial body hair is driven by minute forces external to the skin surface. This structure can be used on any surface of the human body because it is fully composed of passive elements.