The task of Knowlegde-Based Visual Question Answering (KB-VQA) requires the model to understand visual features and retrieve external knowledge. Retrieval-Augmented Generation (RAG) have been employed to address this problem through knowledge base querying. However, existing work demonstrate two limitations: insufficient interactivity during knowledge retrieval and ineffective organization of retrieved information for Visual-Language Model (VLM). To address these challenges, we propose a three-stage visual language model with Process, Retrieve and Filter (VLM-PRF) framework. For interactive retrieval, VLM-PRF uses reinforcement learning (RL) to guide the model to strategically process information via tool-driven operations. For knowledge filtering, our method trains the VLM to transform the raw retrieved information into into task-specific knowledge. With a dual reward as supervisory signals, VLM-PRF successfully enable model to optimize retrieval strategies and answer generation capabilities simultaneously. Experiments on two datasets demonstrate the effectiveness of our framework.

Recent advances in single-cell and spatial transcriptomics have revolutionized our understanding of cellular heterogeneity. However, translating high-dimensional data into functional pathway insights remains challenging. To address this obstacle, we developed PaaSc (Pathway activity analysis of Single-cell), a computational method for inferring pathway activity at single-cell resolution. PaaSc employs multiple correspondence analysis to simultaneously project cells and genes into a common latent space and selects pathway-associated dimensions through linear regression to infer pathway activity scores. We validated PaaSc across diverse benchmarking datasets, including those that jointly profiled protein and RNA levels, as well as large-scale cancer scRNA-seq cohorts. Compared with state-of-the-art methods, PaaSc demonstrated superior performance in multiple applications: scoring cell type-specific gene sets, identifying cell senescence-associated pathways, and exploring GWAS trait-associated cell types. Importantly, PaaSc maintained accuracy despite batch effects and demonstrated robust performance across different data modalities, including scATAC-seq and spatial transcriptomics data. Our results demonstrate that PaaSc accurately captures dynamic cellular states and spatial patterns, thereby advancing our understanding of cellular dynamics, aging, and disease mechanisms.