Search papers, labs, and topics across Lattice.
The paper introduces 3D-VCD, a novel inference-time framework to mitigate hallucinations in 3D embodied agents powered by large multimodal models. It works by constructing distorted 3D scene graphs via semantic and geometric perturbations and then contrasting model predictions under original and distorted contexts to suppress hallucinated tokens. Experiments on 3D-POPE and HEAL benchmarks demonstrate that 3D-VCD improves grounded reasoning without retraining.
Hallucinations in 3D embodied agents can be significantly reduced at inference time by contrasting predictions under original and geometrically/semantically perturbed 3D scene graphs.
Large multimodal models are increasingly used as the reasoning core of embodied agents operating in 3D environments, yet they remain prone to hallucinations that can produce unsafe and ungrounded decisions. Existing inference-time hallucination mitigation methods largely target 2D vision-language settings and do not transfer to embodied 3D reasoning, where failures arise from object presence, spatial layout, and geometric grounding rather than pixel-level inconsistencies. We introduce 3D-VCD, the first inference-time visual contrastive decoding framework for hallucination mitigation in 3D embodied agents. 3D-VCD constructs a distorted 3D scene graph by applying semantic and geometric perturbations to object-centric representations, such as category substitutions and coordinate or extent corruption. By contrasting predictions under the original and distorted 3D contexts, our method suppresses tokens that are insensitive to grounded scene evidence and are therefore likely driven by language priors. We evaluate 3D-VCD on the 3D-POPE and HEAL benchmarks and show that it consistently improves grounded reasoning without any retraining, establishing inference-time contrastive decoding over structured 3D representations as an effective and practical route to more reliable embodied intelligence.
