Introduces the TVR active viewpoint-matching task and TVRBench indoor simulation benchmark, where foundation models start at low single-digit success rates but reach 51.4% after visual-action SFT and multi-turn GRPO post-training.
4 Yanpeng Zhao, Wentao Ding, Hongtao Li, Baoxiong Jia, and Zilong Zheng
6 Pith papers cite this work. Polarity classification is still indexing.
6
Pith papers citing it
years
2026 6representative citing papers
EAGOR reformulates embodied 360-degree directional reasoning as recursive Bayesian estimation on a spherical manifold using spherical harmonics, achieving training-free, rotation-equivariant target tracking.
ESI-Bench shows active exploration outperforms passive observation in multimodal LLMs on spatial tasks but reveals failures from poor action choices and overconfident belief commitment unlike humans.
PanoWorld adds spherical spatial cross-attention and pano-native training data to MLLMs for improved spatial reasoning on ERP panoramas, outperforming baselines on new and existing benchmarks.
Imagining in 360° decouples visual search into a single-step probabilistic semantic layout predictor and an actor, removing the need for multi-turn CoT reasoning and trajectory annotations while improving efficiency in 360° environments.
E3VS-Bench is a 99-scene, 2,014-episode benchmark using 3D Gaussian Splatting to evaluate whether VLMs can actively control 5-DoF viewpoints to gather visual evidence for question answering.
citing papers explorer
-
Where to Look: Can Foundation Models Reach a Target Viewpoint Through Active Exploration?
Introduces the TVR active viewpoint-matching task and TVRBench indoor simulation benchmark, where foundation models start at low single-digit success rates but reach 51.4% after visual-action SFT and multi-turn GRPO post-training.
-
EAGOR: Embodied Reasoning in Omni-direction
EAGOR reformulates embodied 360-degree directional reasoning as recursive Bayesian estimation on a spherical manifold using spherical harmonics, achieving training-free, rotation-equivariant target tracking.
-
ESI-Bench: Towards Embodied Spatial Intelligence that Closes the Perception-Action Loop
ESI-Bench shows active exploration outperforms passive observation in multimodal LLMs on spatial tasks but reveals failures from poor action choices and overconfident belief commitment unlike humans.
-
PanoWorld: Towards Spatial Supersensing in 360$^\circ$ Panorama World
PanoWorld adds spherical spatial cross-attention and pano-native training data to MLLMs for improved spatial reasoning on ERP panoramas, outperforming baselines on new and existing benchmarks.
-
Beyond Thinking: Imagining in 360$^\circ$ for Humanoid Visual Search
Imagining in 360° decouples visual search into a single-step probabilistic semantic layout predictor and an actor, removing the need for multi-turn CoT reasoning and trajectory annotations while improving efficiency in 360° environments.
-
E3VS-Bench: A Benchmark for Viewpoint-Dependent Active Perception in 3D Gaussian Splatting Scenes
E3VS-Bench is a 99-scene, 2,014-episode benchmark using 3D Gaussian Splatting to evaluate whether VLMs can actively control 5-DoF viewpoints to gather visual evidence for question answering.