We report an evaluation benchmark for assessing the operational suitability of Vision-Language Models (VLMs) in fashion e-commerce. General-purpose benchmarks do not adequately cover fashion-specific attributes or the structured extraction tasks common in e-commerce workflows. We define five tasks across two image streams---outfit and single-item product images---and compare six commercial and two open-source models with multiple prompt variants, including a canonical prompt and model-proposed prompts. Experiments show that the best-performing model varies by task, error patterns are more model-dependent than prompt-dependent, and model updates can improve some tasks while degrading others. These results indicate that task-specific evaluation, prompt robustness checks, and continuous monitoring are practical requirements for deploying VLMs in production fashion systems.

Recent research on explainable recommendation generally frames the task as a standard text generation problem, and evaluates models simply based on the textual similarity between the predicted and ground-truth explanations. However, this approach fails to consider one crucial aspect of the systems: whether their outputs accurately reflect the users' (post-purchase) sentiments, i.e., whether and why they would like and/or dislike the recommended items. To shed light on this issue, we introduce new datasets and evaluation methods that focus on the users' sentiments. Specifically, we construct the datasets by explicitly extracting users' positive and negative opinions from their post-purchase reviews using an LLM, and propose to evaluate systems based on whether the generated explanations 1) align well with the users' sentiments, and 2) accurately identify both positive and negative opinions of users on the target items. We benchmark several recent models on our datasets and demonstrate that achieving strong performance on existing metrics does not ensure that the generated explanations align well with the users' sentiments. Lastly, we find that existing models can provide more sentiment-aware explanations when the users' (predicted) ratings for the target items are directly fed into the models as input. The datasets and benchmark implementation are available at: https://github.com/jchanxtarov/sent_xrec.