Personality alignment tailors Large Language Models (LLMs) for personalized interactions. Most existing studies focus on prompting, post- training, or representation steering to elicit trait expression, but they often compromise the general capabilities like reasoning and truthfulness. In this work, we empirically verified the existence of the Personality Alignment Tax (PAT) that distribution shifts and conflicting objectives during personality updates overwrite pre-trained competencies. To mitigate this tax, we propose a Personality-Orthogonal Gradient Projection (POGP) plug-in that preserves the capability in personality alignment. Specifically, POGP estimates the general-capability subspace from capability reference benchmarks and projects personality gradients onto the orthogonal complement before updating. This isolates and neutralizes conflicting gradients while maintaining the pre-trained reasoning capabilities. Experiments on the BIG5-CHAT dataset with two backbone LLMs demonstrate that POGP achieves a stronger personality plasticity and capability stability trade-off than baselines.

Recent advances in Vision-Language Models (VLMs) have achieved impressive progress in multimodal mathematical reasoning. Yet, howmuch visual information truly contributes to reasoning remains unclear. Existing benchmarks report strong overall performance but seldom isolate the role of the image modality, leaving open whether VLMs genuinely leverage visual understanding or merely depend on linguistic priors. To address this, we present MathSight, a university-level multimodal mathematical reasoning benchmark designed to disentangle and quantify the effect of visual input. Each problem includes multiple visual variants—original, hand-drawn, photocaptured—and a text-only condition for controlled comparison. Experiments on state-of-the-art VLMs reveal a consistent trend: the contribution of visual information diminishes with increasing problem difficulty. Remarkably, Qwen3-VL without any image input surpasses both its multimodal variants and GPT-5, underscoring the need for benchmarks like MathSight to advance genuine vision-grounded reasoning in future models. The project page is available at https://cnu-bot-group.github.io/MathSight/.