Enhancing Part-Level Point Grounding for Any Open-Source MLLMs
About
Visual grounding aims to associate free-form textual queries with specific regions in an image. While recent Multimodal Large Language Models (MLLMs) have demonstrated promising capabilities in this domain, they primarily excel at object-level grounding and often struggle with part-level grounding-an essential requirement for fine-grained tasks such as robotic manipulation. In this work, we introduce a general approach that equips any open-source MLLMs with accurate 2D part-level point grounding, offering a more direct alternative to conventional grounding representations. Our method leverages the attention mechanisms inherently present in MLLMs. By synthesizing text-conditioned, grounding-aware queries within intermediate layers via the proposed Q-Synth Module, we capture target-relevant attention patterns and refine them with a lightweight Attention-to-Point Decoder, which converts these patterns into a point-centric heatmap for final prediction. Notably, all original MLLM parameters are frozen, ensuring full preservation of their pre-trained capabilities. Experiments show that our design consistently improves part-level grounding accuracy across datasets and can be seamlessly integrated into any open-source MLLMs.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Point Grounding | PACO (test) | Patch Accuracy61 | 24 | |
| Point Grounding | InstructPart (test) | Patch Accuracy90 | 13 | |
| Reasoning Pointing | InstructPart (test) | Patch Accuracy90 | 11 | |
| Point Grounding | PointArena Point-Bench (cross-dataset evaluation) | Affordance P-Acc83.8 | 6 | |
| Object-level point grounding | PACO 19 (test) | -- | 3 | |
| Part-level point grounding | PACO 19 (test) | -- | 3 |