Language-guided semantic cues from MLLM visual pipelines, steered by text embeddings, refine object semantics and boost grounding accuracy against occlusion and small objects.
Clip-convnextlarge d 320.laion2b-s29b-b131k- ft-soup,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Robust Grounding with MLLMs Against Occlusion and Small Objects via Language-Guided Semantic Cues
Language-guided semantic cues from MLLM visual pipelines, steered by text embeddings, refine object semantics and boost grounding accuracy against occlusion and small objects.