arXiv subjects

Yonghao Shi

Publications and source records attributed to Yonghao Shi.

3 recordsLinked to original sources

arXiv2026

HEARTH: An Object-Centric RGB-Thermal-3D Dataset for Temperature-Aware Robot Manipulation

Language-guided manipulation can depend on physical properties that visible appearance does not reveal. Temperature is one such property, but object datasets for robot learning rarely associate measured temperatures with object appearance and geometry. We present HEARTH, an object-centric RGB-thermal-3D dataset of 90 physical objects from 18 everyday categories, comprising 145 captured object states. Our pipeline maps apparent surface temperatures onto reconstructed meshes through camera calibration and pose transfer. The dataset includes raw temperature measurements, camera parameters, RGB-textured meshes, and thermal textures for simulation. We use these assets to construct three LIBERO-derived tasks and collect 1,200 demonstrations for fine-tuning a pretrained vision-language-action (VLA) model, $π_{0.5}$. In an ablation study, adding thermal observations to the VLA increases success on temperature-dependent object-selection tasks from 35.0% for the RGB-only baseline to 75.0%. These results demonstrate the utility of HEARTH for training robot policies to follow temperature-related instructions.

cs.RO↗

arXiv2025

Step-level Verifier-guided Hybrid Test-Time Scaling for Large Language Models

Test-Time Scaling (TTS) is a promising approach to progressively elicit the model's intelligence during inference. Recently, training-based TTS methods, such as continued reinforcement learning (RL), have further surged in popularity, while training-free TTS methods are gradually fading from prominence. However, the additional computation overhead of training amplifies the burden on test-time scaling. In this paper, we focus on training-free TTS methods for reasoning. We first design Conditional Step-level Self-refinement, a fine-grained sequential scaling method guided by process verification. On top of its effectiveness, we further combine it with other classical parallel scaling methods at the step level, to introduce a novel inference paradigm called Hybrid Test-Time Scaling. Extensive experiments on five instruction-tuned LLMs across different scales (3B-14B) and families demonstrate that hybrid strategy incorporating various training-free TTS methods at a fine granularity has considerable potential for expanding the reasoning performance boundaries of LLMs.

cs.CL↗

arXiv2021

SoMIAP: Social media images analysis and prediction framework

The personal photos captured and submitted by users on social networks can provide several interesting insights about the location of the user, which is a key indicator of their daily activities. This information is invaluable for security organisations, especially for security monitoring and tracking criminal activities. Hence, we propose in this paper a novel approach for location prediction based on the image analysis of the photos posted on social media. Our approach combines two main methods to perform the image analysis, place and face recognition. The first method is used to determine the location area in the analysed image. The second is used to identify people in the analysed image, by locating a face in the image and comparing it with a dataset of images that have been collected from different social platforms. The effectiveness of the proposed approach is demonstrated through performance analysis and experimental results.

cs.SI↗