2 papers
cs.AI2026
RAGStress: A controlled benchmark for evaluating retrieval-augmented generation under knowledge-base degradation
Shiqi Yang, Jiekai Ma, Gaoyuan Du
Retrieval-Augmented Generation (RAG) is typically evaluated under the implicit assumption that the underlying knowledge base (KB) is clean, leaving the behaviour of RAG systems und…
cs.CV2026
Beyond Plausibility: Verifiable Fine-Grained Image Editing on Structured Assets
Muyao Wang, Chen Zhu, Shiqi Yang +2
Fine-grained image editing requires more than producing a visually plausible result: an editor must execute the requested attribute change precisely while leaving everything else i…