Large language models (LLMs) make it possible to build knowledge graphs from professional literature at scale, but how such graphs fail at that scale is rarely measured. We built \sys{}, a cross-media knowledge graph of swine, aquaculture, tomato, maize and rice production, and analysed its 199,418 entities and 277,053 relations together with the web system that serves it. An LLM extracted typed triples from the professional literature of each sector; 74.9% of the relations carry an evidence record of document, page and excerpt, and a multimodal embedding model links 25,158 entities to images and 14,132 to video clips. We identified five failure modes of construction. First, the sector schem... More >

Full-cycle management of greenhouse tomato requires continuous information on crop state together with estimates of daily water and nutrient use, yet destructive biomass measurements are sparse, and a single fixed model shared across resources is prone to stage-dependent negative transfer. This study proposes Dual-Path Multi-timescale Fusion (DPMF), a development-aware framework for reconstructing crop state from sparse observations and estimating daily resource fluxes. The state path combines a process-based dry-matter prior with a gated recurrent unit trained on three kinds of supervision: point states, biomass increments between adjacent samples, and cumulative harvested-fruit dry matter.... More >