Large-scale deep learning models with a pretraining-finetuning paradigm have led to a surge of numerous task-specific models fine-tuned from a common pre-trained model. Recently, several research efforts have been made on merging these large models into a single multi-task model, particularly with simple arithmetic on parameters. Such merging methodology faces a central challenge: interference between model parameters fine-tuned on different tasks. Few recent works have focused on designing a new fine-tuning scheme that can lead to small parameter interference, however at the cost of the performance of each task-specific fine-tuned model and thereby limiting that of a merged model. To improve the performance of a merged model, we note that a fine-tuning scheme should aim for (1) smaller parameter interference and (2) better performance of each fine-tuned model on the corresponding task. In this work, we aim to design a new fine-tuning objective function to work towards these two goals. In the course of this process, we find such objective function to be strikingly similar to sharpness-aware minimization (SAM) objective function, which aims to achieve generalization by finding flat minima. Drawing upon our observation, we propose to fine-tune pre-trained models via sharpness-aware minimization. The experimental and theoretical results showcase the effectiveness and orthogonality of our proposed approach, improving performance upon various merging and fine-tuning methods. Our code is available at https://github.com/baiklab/SAFT-Merge.

Recently, self-tracking devices such as wearable activity trackers have become more available to end users. While these emerging products are imbued with new characteristics in terms of human-computer interaction, it is still unclear how to describe and design for user experience in such devices. In this paper, we present a three-week field study, which aimed to unfold users' experience with wearable activity trackers. Drawing from Knapp's model of interaction stages in interpersonal relationship development, we propose three stages of relationship development between users and self-tracking devices: initiation & experimentation, intensifying & integration, and stagnation & termination. We highlight the challenges in each stage and design opportunities for future self-tracking devices.

People become increasingly influenced by others in changing and maintaining health behaviors. Along with the advancement of persuasive technology and social networking technologies, the place where social interaction occurs has expanded. As a result, mediators who influence an individual's behavior change can come from diverse social channels. However, little work exists on what roles the mediators have and how differently the mediators motivate and affect the maintenance of health behavior changes of users through various social channels. To investigate this, we conducted interviews with 13 participants who use a running exercise application for maintaining their health behavior changes. This study reveals the roles of mediators from three different social channels, which are the social feature in the application, general social media, and the agent feature in the application. Mediators from the application could influence participants' health behavior change either positively or negatively according to the level of intimacy and the similarity of the physical condition. Social media mediators influence participants' social face and support their health behavior changes by keeping participants in countenance. Lastly, the agent mediator of the application provides continuous reinforcement to participants for maintaining their health behavior changes.

People worldwide are increasingly acquiring collections of virtual possessions. While virtual possessions have become ubiquitous, little work exists on how people value and form attachments to these things. To investigate, we conducted a study with 48 young adults from South Korea, Spain and the United States. The study probed on participants' perceived value of their virtual possessions as compared to their material things, and the comparative similarities and differences across cultures. Findings show that young adults live in unfinished spaces and that they often experience a sense of fragmentation when trying to integrate their virtual possessions into their lives. These findings point to several design opportunities, such as tools for life story-oriented archiving, and insights on better forms of Cloud storage.