Researchers have introduced Shift & Drift, a new benchmark designed to evaluate the robustness and generalization of autonomous driving motion planners under distribution shifts.
A large-scale study analyzes the usage patterns of an AI-based learning assistant among over 77,000 higher education students in distance learning programs.
A new evaluation suite called Blind-Spots-Bench has been developed to identify simple tasks that humans easily perform but multimodal AI models struggle to complete.
The AutoPersonas framework addresses the issue of behavioral collapse in long-term persona agents by enabling open-ended evolution across multiple timescales.
The authors present Diffusion-GR2, a generative reasoning re-ranker that utilizes block-diffusion language models to accelerate inference speeds in recommendation systems.
Researchers developed CatDT, a self-evolving multi-agent digital twin system designed to autonomously simulate and discover heterogeneous catalysts.
MasFACT is a continual learning framework designed to optimize and adapt communication topologies among multi-agent systems across evolving tasks.
A prompting framework called Concretized Proposition Prompting has been developed to improve reasoning and knowledge retrieval in large language models, particularly for medical tasks.
A new behavioral benchmark evaluates how closely the emergent object segmentation properties of self-supervised vision transformers align with human visual perception.
This paper introduces an online, reference-free evaluation framework to monitor the quality of flowchart image-to-code generation by vision-language models during inference.
The MultiFair framework addresses modality bias and demographic unfairness in multimodal medical classification models using dual-level gradient modulation.
Researchers have introduced InvestPhilBench, a multi-layer benchmark designed to evaluate how well large language models can reconstruct and apply expert investment decision frameworks.
Track2Map is a new real-time 3D Gaussian Splatting system designed to reconstruct deformable anatomy during robotic surgery without relying on precise camera trajectory priors.
The proposed LEEVLA architecture improves vision-language-action models for robotics by focusing on critical environmental changes rather than treating all visual data equally.
This survey paper categorizes and reviews recent system-level optimizations for key-value caches to improve the efficiency and throughput of large language model serving.
A new graph-based framework evaluates the logical consistency and uncertainty of large language model reasoning steps rather than just checking the final output.
A study evaluates the validity of Portugal's AMALIA language model when used as an automated data annotator for identifying moral authority.
This paper explores optimal teaching strategies for inverse reinforcement learning to help autonomous agents learn robust reward functions across multiple environments and modalities.
The study benchmarks the performance of robotic diffusion policies as their observation context length is increased, addressing memory limitations in dexterous manipulation.
A new dual-system framework called FSD-VLN improves aerial vision-language navigation by separating high-level semantic reasoning from low-latency flight control.
Wanyun Technology has launched an automated charging service engine designed to handle complex, real-world charging scenarios across public and private sectors.
Bloomage Biotech has established a smart manufacturing subsidiary in Tianjin with a registered capital of 10 million RMB to focus on cosmetics, pharmaceuticals, and medical device production.
Galaxy Securities recommends that investors focus on financial performance and opportunities within the semiconductor supply chain, including memory chips and advanced packaging, amid stock market fluctuations.
Heavy-lift drones were deployed in Guangxi, China, to airlift flood victims, highlighting the growing capabilities of the country's low-altitude aviation sector.