Generating charts that accurately reflect complex data remains a subtle challenge in today's data visualization landscape. Charts require not only precise layout, color, and text placement, but also the translation of these visual details into code to reproduce the intended design. However, traditional methods often rely on directly prompting vision-language models (VLMs), such as GPT-4V, which frequently struggle to translate complex visual elements into grammatically correct Python code. Even minor errors can result in charts failing to meet their design objectives.

A collaborative team from Ant Group, Zhejiang University, University of Liverpool, and East China Normal University presented ScaleOT, a novel cross-domain fine-tuning framework, at AAAI 2025. This framework achieves a 50% improvement in privacy protection and a remarkable 90% reduction in computational cost compared to knowledge distillation, without compromising model performance. ScaleOT offers a highly efficient and lightweight solution for cross-domain fine-tuning of billion-parameter models. The innovative nature of the paper secured it an oral presentation at AAAI.

In the field of artificial intelligence (AI), while large language models (LLMs) excel at processing natural language, they often struggle with complex reasoning tasks. These tasks typically require multi-step reasoning, domain-specific knowledge, or effective integration of external tools. To overcome these limitations, researchers have been exploring how to enhance LLM capabilities through the use of external tools. Traditional enhancement methods usually necessitate fine-tuning or additional training of the models, which limits their adaptability and flexibility for tasks.

Recently, an interdisciplinary research team from the Technical University of Munich, the Helmholtz Zentrum Munich, and ETH Zurich published an important study introducing an innovative framework called Moscot (Multi-Omics Single-Cell Optimal Transport), which successfully reconstructed the developmental trajectories of 1.7 million mouse embryonic cells at 20 time points. This study, published in the journal Nature, marks a significant breakthrough in the field of single-cell genomics. The design inspiration for the Moscot framework comes from the optimal transport theory of the 18th century.

With the advancement of artificial intelligence technology, multi-agent systems are increasingly capable of handling complex tasks across various fields. These systems are composed of multiple specialized agents that collaborate to leverage their individual strengths and achieve common goals. Such collaboration excels in areas like complex reasoning, programming, drug discovery, and safety assurance, as the structured interactions between agents not only enhance problem-solving efficiency but also allow for mutual correction, thereby improving each agent's output. Research indicates that this collaborative approach often outperforms in tasks that require rigorous reasoning or fact verification.

Meta has recently launched a new risk policy framework aimed at assessing and mitigating the risks associated with cutting-edge AI models, with the intention to halt development or restrict the release of these systems when necessary. This framework, named 'Cutting-edge AI Framework', outlines how Meta will categorize AI models into high-risk and critical risk categories, and take appropriate measures to reduce risks to an 'acceptable level'. In this framework, critical risks are defined as those uniquely capable of executing specific threat scenarios, while high risk implies

A recent study released by Apple reveals that the company is developing a new framework called ELEGNT, aimed at enabling non-anthropomorphic robots to exhibit more natural and expressive movements, which could open new avenues for the future development of home robots. The research team used a robot modeled after the lamp character in Pixar's animation "Luxo Jr.", equipped with a 6-axis robotic arm and a head featuring lighting and projection capabilities. Through the ELEGNT framework, researchers have enabled the robot to not only perform functional tasks but also convey intent and emotion through its movements.

In the field of academic research, literature retrieval is a complex and crucial information acquisition task. Researchers need the capability to handle intricate, domain-specific search queries to meet detailed research requirements. However, existing academic search platforms, like Google Scholar, often struggle to address these complex research queries. For example, queries related to non-stationary reinforcement learning using the UCB method require stronger computational and analytical capabilities. Moreover, researchers typically need to spend a significant amount of time and effort manually browsing extensive academic literature when conducting literature reviews.

A research team from New York University, MIT, and Google has recently proposed an innovative framework aimed at addressing the bottleneck of inference time scaling in diffusion models. This groundbreaking research transcends the traditional method of simply increasing the denoising steps, opening up new avenues for improving generative model performance. The framework primarily unfolds from two dimensions: one is leveraging feedback from validators, and the second is implementing algorithms to discover better noise candidates. The research team builds on a pre-trained SiT-XL model with a resolution of 256×256, maintaining 250 fixed denoising steps.