SIN-Bench: Tracing Native Evidence Chains in Long-Context Multimodal Scientific Interleaved Literature

SIN-Bench: Tracing Native Evidence Chains in Long-Context Multimodal Scientific Interleaved Literature explores Develop tools to enhance the evidence-based reasoning capabilities of multimodal language models in scientific literature.. Commercial viability score: 4/10 in Multimodal Scientific Analysis.