GaussianWAM: Distilling Geometry and Semantics from 3D Gaussian Fields into World-Action Models
本文提出GaussianWAM,通过3D高斯场整合几何和语义监督信息以增强WAM的表示学习,从而提高机器人操作中的视觉预测与动作生成性能。
本文提出GaussianWAM,通过3D高斯场整合几何和语义监督信息以增强WAM的表示学习,从而提高机器人操作中的视觉预测与动作生成性能。
本文提出ViTexSZ框架,通过视觉-文本知识蒸馏方法解决异构EEG信号癫痫检测问题,提高检测准确性和泛化能力。
It remains unclear how to effectively construct small language models (SLMs) that exhibit high trustworthiness across multiple dimensions—fairness, robustness, privacy, and ethics. This work presents the first systematic comparison between training SLMs from scratch and compressing large language models (LLMs), introducing a comprehensive evaluation framework that assesses trustworthiness along these four axes. The study investigates the impact of pruning, quantization, and knowledge distillation on the trust-related properties of SLMs. Findings reveal that quantization preserves trustworthiness significantly better than pruning. Moreover, SLMs derived via quantization from trustworthy LLMs outperform natively trained small models in both trustworthiness and task adaptability. Further gains in reliability are achievable by incorporating knowledge distillation into the compression pipeline.
This work addresses the challenge of formal verification for Reflex programs in industrial-scale control systems, where the generation of an excessive number of verification conditions often renders manual analysis impractical. To overcome this limitation, the authors propose a hybrid verification strategy that integrates a structured requirement annotation language with automated invariant inference based on program structure, coupled with an SMT solver to automatically discharge a substantial subset of verification conditions. By leveraging this synergistic approach, the method significantly reduces the number of verification tasks requiring human intervention, thereby enhancing the automation, feasibility, and overall efficiency of formal verification for large-scale process control systems.
This work addresses the high cost and risk of real-world reinforcement learning for autonomous driving, where existing pixel-level diffusion-based world models suffer from prohibitive inference latency (~2 seconds per frame), hindering high-frequency interaction. To overcome this, the authors propose DreamerAD, a latent-space world model featuring three key innovations: shortcut forcing via recursive multi-resolution step compression, a latent-representation-based autoregressive dense reward model, and Gaussian vocabulary sampling tailored for GRPO. These mechanisms collectively reduce diffusion sampling from 100 steps to a single step—yielding an 80× speedup—while preserving visual interpretability. Evaluated on NavSim v2, DreamerAD achieves a state-of-the-art 87.7 EPDMS, establishing a new performance benchmark and demonstrating the efficacy and practicality of latent-space reinforcement learning for autonomous driving.
本文提出GaussianWAM,通过3D高斯场整合几何和语义监督信息以增强WAM的表示学习,从而提高机器人操作中的视觉预测与动作生成性能。
本文提出ViTexSZ框架,通过视觉-文本知识蒸馏方法解决异构EEG信号癫痫检测问题,提高检测准确性和泛化能力。
It remains unclear how to effectively construct small language models (SLMs) that exhibit high trustworthiness across multiple dimensions—fairness, robustness, privacy, and ethics. This work presents the first systematic comparison between training SLMs from scratch and compressing large language models (LLMs), introducing a comprehensive evaluation framework that assesses trustworthiness along these four axes. The study investigates the impact of pruning, quantization, and knowledge distillation on the trust-related properties of SLMs. Findings reveal that quantization preserves trustworthiness significantly better than pruning. Moreover, SLMs derived via quantization from trustworthy LLMs outperform natively trained small models in both trustworthiness and task adaptability. Further gains in reliability are achievable by incorporating knowledge distillation into the compression pipeline.
This work addresses the challenge of formal verification for Reflex programs in industrial-scale control systems, where the generation of an excessive number of verification conditions often renders manual analysis impractical. To overcome this limitation, the authors propose a hybrid verification strategy that integrates a structured requirement annotation language with automated invariant inference based on program structure, coupled with an SMT solver to automatically discharge a substantial subset of verification conditions. By leveraging this synergistic approach, the method significantly reduces the number of verification tasks requiring human intervention, thereby enhancing the automation, feasibility, and overall efficiency of formal verification for large-scale process control systems.
This work addresses the high cost and risk of real-world reinforcement learning for autonomous driving, where existing pixel-level diffusion-based world models suffer from prohibitive inference latency (~2 seconds per frame), hindering high-frequency interaction. To overcome this, the authors propose DreamerAD, a latent-space world model featuring three key innovations: shortcut forcing via recursive multi-resolution step compression, a latent-representation-based autoregressive dense reward model, and Gaussian vocabulary sampling tailored for GRPO. These mechanisms collectively reduce diffusion sampling from 100 steps to a single step—yielding an 80× speedup—while preserving visual interpretability. Evaluated on NavSim v2, DreamerAD achieves a state-of-the-art 87.7 EPDMS, establishing a new performance benchmark and demonstrating the efficacy and practicality of latent-space reinforcement learning for autonomous driving.