Selected work

Representative Papers

SCALE: State-Calibrated Latent Embeddings for JEPA Planning in the Right Geometry

Aug 17, 2026

This study addresses the failure of state guidance in JEPA planning caused by suboptimal representational geometry. We propose SCALE, a method that aligns the latent space with the task state space via pairwise distance correlation regularization to achieve state-calibrated embeddings, thereby enabling state information to effectively dominate planning cost computation. Our findings reveal that planning performance depends critically on representational geometry rather than merely information presence. Experiments demonstrate that SCALE consistently outperforms LeWM across five tasks and multiple solvers without additional inference overhead. Consequently, this work establishes a novel paradigm for lightweight optimization in JEPA-based planning by ensuring geometrically faithful state representations.

0 citationsRead paper

3D Primitives are a Spatial Language for VLMs

May 12, 2026

This work addresses the limited spatial reasoning capabilities of vision-language models (VLMs), which, despite generating 3D primitive code containing object categories, counts, and coarse locations, often fail to achieve accurate spatial understanding. To bridge this gap, the authors propose using executable 3D geometric primitive code as an intermediate representation for spatial reasoning and introduce three key contributions: the SpatialBabel benchmark to evaluate the impact of multilingual scene code on VLM performance, a training-free Code-CoT reasoning strategy, and an unsupervised S³-FT fine-tuning method that leverages model-generated primitives for knowledge distillation. Experiments demonstrate that Code-CoT improves performance by 6.4% and 5.0% on SpatialBabel-QA and CV-Bench-3D, respectively, while S³-FT boosts Qwen3-VL-8B by 4.6%–17% across multiple benchmarks, with the approach showing strong cross-model generalizability.

0 citationsRead paper

InfoFusion Controller: Informed TRRT Star with Mutual Information based on Fusion of Pure Pursuit and MPC for Enhanced Path Planning

Mar 08, 2025

To address high uncertainty and insufficient obstacle-avoidance robustness in local path planning for autonomous vehicles operating in complex, dynamic urban environments with unpredictable obstacles, this paper proposes a global–local cooperative path planning framework. At the global level, Informed-TRRT* ensures asymptotically optimal path search; at the local level, model predictive control (MPC) is fused with a pure pursuit controller, where mutual information (MI) is introduced—novelly—for dynamic, adaptive weighting of the two controllers. The framework jointly optimizes global path optimality and local real-time responsiveness, and natively supports integration with SLAM-based mapping. Extensive experiments across multiple complex urban maps demonstrate that our method reduces collision rate by 42%, decreases lateral tracking error by 37%, and lowers average planning latency by 29% compared to baseline approaches. The source code is publicly available.

0 citationsRead paper

Recent publications

Latest Papers

SCALE: State-Calibrated Latent Embeddings for JEPA Planning in the Right Geometry

Aug 17, 2026

This study addresses the failure of state guidance in JEPA planning caused by suboptimal representational geometry. We propose SCALE, a method that aligns the latent space with the task state space via pairwise distance correlation regularization to achieve state-calibrated embeddings, thereby enabling state information to effectively dominate planning cost computation. Our findings reveal that planning performance depends critically on representational geometry rather than merely information presence. Experiments demonstrate that SCALE consistently outperforms LeWM across five tasks and multiple solvers without additional inference overhead. Consequently, this work establishes a novel paradigm for lightweight optimization in JEPA-based planning by ensuring geometrically faithful state representations.

0 citationsRead paper

3D Primitives are a Spatial Language for VLMs

May 12, 2026

This work addresses the limited spatial reasoning capabilities of vision-language models (VLMs), which, despite generating 3D primitive code containing object categories, counts, and coarse locations, often fail to achieve accurate spatial understanding. To bridge this gap, the authors propose using executable 3D geometric primitive code as an intermediate representation for spatial reasoning and introduce three key contributions: the SpatialBabel benchmark to evaluate the impact of multilingual scene code on VLM performance, a training-free Code-CoT reasoning strategy, and an unsupervised S³-FT fine-tuning method that leverages model-generated primitives for knowledge distillation. Experiments demonstrate that Code-CoT improves performance by 6.4% and 5.0% on SpatialBabel-QA and CV-Bench-3D, respectively, while S³-FT boosts Qwen3-VL-8B by 4.6%–17% across multiple benchmarks, with the approach showing strong cross-model generalizability.

0 citationsRead paper

InfoFusion Controller: Informed TRRT Star with Mutual Information based on Fusion of Pure Pursuit and MPC for Enhanced Path Planning

Mar 08, 2025

To address high uncertainty and insufficient obstacle-avoidance robustness in local path planning for autonomous vehicles operating in complex, dynamic urban environments with unpredictable obstacles, this paper proposes a global–local cooperative path planning framework. At the global level, Informed-TRRT* ensures asymptotically optimal path search; at the local level, model predictive control (MPC) is fused with a pure pursuit controller, where mutual information (MI) is introduced—novelly—for dynamic, adaptive weighting of the two controllers. The framework jointly optimizes global path optimality and local real-time responsiveness, and natively supports integration with SLAM-based mapping. Extensive experiments across multiple complex urban maps demonstrate that our method reduces collision rate by 42%, decreases lateral tracking error by 37%, and lowers average planning latency by 29% compared to baseline approaches. The source code is publicly available.

0 citationsRead paper