#LLM as a judge

1 stories · 1 in 7 days · 1 in 30 days
Stories, products and papers tagged "LLM as a judge", newest first.
Nov 18
How-toBuildersSelected03:04
LlamaIndex founder explains granular design in ParseBench-2 and DPP

The post reveals that benchmark scoring for document parsing is shifting from single aggregate scores toward severity-weighted, dimension-level evaluation. Teams building RAG pipelines or data extraction workflows can use ParseBench-2's scenario set and DPP's judge prompts directly to customise evaluation against the error types that hurt their use case. Compared with coarse end-to-end parsing scores, this approach lets a wrong-digit failure outweigh cosmetic markup noise, and both resources are openly accessible.

Jerry Liu@jerryjliu0Source