We present SWIM (See What I Mean), a novel training strategy that aligns vision and language representations to enable fine-grained object understanding solely from textual prompts. Unlike existing approaches that require explicit visual prompts, such as masks or points, SWIM leverages mask supervision only during training to guide cross-modal attention, allowing the model to automatically attend to the user-specified object at inference. Our cross-attention analysis of pretrained multimodal large languagemodels (MLLMs) reveals a systematic discrepancy: Attribute words produce sharp, localized activations in the visual modality, whereas object nouns yield diffuse and scattered patterns due to semantic reference bias and distributed high-level representations. To address this misalignment, we construct NL-Refer, an enriched dataset, in which each object mask is paired with a precise natural language referring expression. SWIM extracts multi-layer cross-attention maps from object nouns and enforces spatial consistency with ground-truth masks. Experimental results demonstrate that SWIM substantially improves text-visual alignment and achieves superior performance over visual-prompt-based methods on fine-grained object understanding benchmarks. The code and data are available at \href{https://github.com/HumanMLLM/SWIM}{https://github.com/HumanMLLM/SWIM}.

翻译：[translated abstract in Chinese] 我们提出SWIM（See What I Mean）——一种新型训练策略，通过对齐视觉与语言表征，实现仅凭文本提示即可完成细粒度对象理解。与需要显式视觉提示（如掩码或点）的现有方法不同，SWIM仅在训练阶段利用掩码监督来引导跨模态注意力机制，使模型在推理时能自动关注用户指定的对象。我们对预训练多模态大语言模型（MLLMs）的交叉注意力分析揭示出系统性偏差：属性词在视觉模态中产生尖锐的局部激活，而对象名词则因语义参考偏差和分布式高层表征呈现弥散化模式。为解决这种错位，我们构建了富化数据集NL-Refer，其中每个对象掩码均配有精确的自然语言指代表达式。SWIM从对象名词中提取多层交叉注意力图，并强制其与真实掩码保持空间一致性。实验表明，SWIM显著提升了文本-视觉对齐性能，在细粒度对象理解基准测试中超越了基于视觉提示的方法。代码与数据已开源至 \href{https://github.com/HumanMLLM/SWIM}{https://github.com/HumanMLLM/SWIM}。