Despite being trained on vast amounts of data, most LLMs are unable to reliably generate well-designed UIs. Designer feedback is essential to improving performance on UI generation; however, we find that existing RLHF methods based on ratings or rankings are not well-aligned with with designers' workflows and ignore the rich rationale used to critique and improve UI designs. In this paper, we investigate several approaches for designers to give feedback to UI generation models, using familiar interactions such as commenting, sketching and direct manipulation. We first perform an evaluation with 21 designers where they gave feedback using these interactions, which resulted in 1500 design annotations. We then use this data to finetune a series of LLMs to generate higher quality UIs. Finally, we evaluate these models with human judges, and we find that our designer-aligned approaches outperform models trained with traditional ranking feedback and all tested baselines, including GPT-5.

Recent research has begun to examine the potential of automatically finding and fixing accessibility issues that manifest in software. However, while recent work makes important progress, it has generally been skewed toward identifying issues that affect users with certain disabilities, such as those with visual or hearing impairments. However there are other groups of users with different types of disabilities that also need software tooling support to improve their experience. As such, this paper aims to automatically identify accessibility issues that affect users with motor-impairments.

Cross-domain recommendation (CDR) has recently emerged as an effective way to alleviate the cold-start and sparsity issues faced by recommender systems, by transferring information from an auxiliary domain to a target domain to improve recommendations. Studying the similarity between domains is a novel direction in CDR research, potentially opening doors for further exploration. In this context, we introduce a systematic approach to quantify similarity between a pair of domains and explore how current CDR methods perform with both similar and dissimilar domain combinations. We achieve this by presenting two original similarity metrics. Our extensive empirical evaluation on different domain combinations demonstrates that the state-of-the-art CDR algorithms do not perform significantly better when using source domains that are more similar to the target domain, compared to those that are less similar. Importantly, we find that no matter how similarity is measured, it does not correlate with the recommendation performance of the state-of-the-art algorithms.

This paper discusses a research agenda at the intersection of Machine Learning, Software Engineering, and Human-Computer Interaction aimed at creating intelligent tools that enable developers to build more accessible software. First, we discuss the creation of MOTOREASE, a novel approach that utilizes computer vision and text processing techniques to identify accessibility issues in mobile app UIs for motor-impaired users. The tool detects four motor-impaired user-focused UI design guidelines: touch target size, expanding sections, persisting elements, and adjacent icon distance, with an average accuracy of around 90%. This represents a significant step towards improving software accessibility for all users. Finally, we discuss our future work for better-supporting developers with tools to engineer accessible software.