Reproduce “IntReal: A Real-World Dataset and Benchmark for Multimodal Interleaved Generation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shaohan Yu, Jie Qin, Wende Tan, Haiwen Diao, Limeng Qiao, Yueming Lyu, Chenyang Si
IntReal pairs a taxonomy-guided real-world dataset with a benchmark for interleaved text-and-image generation. It evaluates complete responses and next-image generation from fixed contexts across seven capability-oriented metrics. For SenseNova-U1-8B-MoT, the seven-metric mean rises from 71.88 after supervised fine-tuning to 73.81 after image-focused reinforcement learning under interleaved contexts, versus 72.31 after fine-tuning alone; this sequential comparison does not isolate reinforcement learning.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.