WebRISE benchmark evaluates MLLM-generated web artifacts

Arxiv CS.CL (NLP/language models)

WebRISE compiles task requirements into states and transitions to assess correctness of MLLM-generated web pages. Unlike existing benchmarks, it captures requirement-induced behavior beyond local evidence.

benchmarks

Arxiv CS.CL (NLP/language models)·Yuxin MengYuhan SuoJunjie WangYuhan SunYiyao YuRuixu ZhangRuining HuYubin WangShouwei RuanBin WangYuxiang ZhangYujiu Yang

8 days ago