Multimodal large language models (MLLMs) have shown impressive capabilities in vision-language tasks such as reasoning segmentation, where models generate segmentation masks based on textual queries. While prior work has primarily focused on perturbing image inputs, semantically equivalent textual paraphrases—crucial in real-world applications where users express the same intent in varied ways—remain underexplored. To address this gap, we introduce a novel adversarial paraphrasing task: generating grammatically correct paraphrases that preserve the original query meaning while degrading segmentation performance. To evaluate the quality of adversarial paraphrases, we develop a comprehensive automatic evaluation protocol validated with human studies. Furthermore, we introduce SPARTA—a black-box, sentence-level optimization method that operates in the low-dimensional semantic latent space of a text autoencoder, guided by reinforcement learning. SPARTA achieves significantly higher success rates, outperforming prior methods by up to 2x on both the ReasonSeg and LLMSeg-40k datasets. We use SPARTA and competitive baselines to assess the robustness of advanced reasoning segmentation models. We reveal that they remain vulnerable to adversarial paraphrasing—even under strict semantic and grammatical constraints. All code and data will be released publicly upon acceptance.

Vision Language Action (VLA) models are widely used in Embodied AI, enabling robots to interpret and execute language instructions. However, their robustness to natural language variability in real-world scenarios has not been thoroughly investigated.In this work, we present a novel systematic study of the robustness of state-of-the-art VLA models under linguistic perturbations. Specifically, we evaluate model performance under two types of instruction noise: (1) human-generated paraphrasing and (2) the addition of irrelevant context. We further categorize irrelevant contexts into two groups according to their length and their semantic and lexical proximity to robot commands. In this study, we observe consistent performance degradation as context size expands. We also demonstrate that the model can exhibit relative robustness to random context, with a performance drop within 10%, while semantically and lexically similar context of the same length can trigger a quality decline of around 50%. Human paraphrases of instructions lead to a drop of nearly 20%. Our results highlights a critical gap in the safety and efficiency of modern VLA models for real-world deployment.

Playlist generation based on textual queries using large language models (LLMs) is becoming an important interaction paradigm for music streaming platforms. User queries span a wide spectrum from highly personalized intent to essentially catalog-style requests. Existing systems typically rely on non-personalized retrieval/ranking or apply a fixed level of preference conditioning to every query, which can overfit catalog queries to a single user or under-personalize explicitly listener-dependent requests. We present an industrial-scale LLM-based playlist generation system with dynamic personalization that adapts the personalization strength to the query type. We define a query taxonomy, train a query-type classifier on 5,000 manually labeled queries, and use its predicted probability to modulate the mixture of LLM-based semantic scoring and personalized evaluation. In a blind user study with pairwise comparisons and ELO aggregation, this approach consistently outperforms both non-personalized and fixed-personalization baselines.