See the main Decisions Explorer page before reading further.
This page is based on the following:

  1. Cabinet decision texts are processed with llama3 via langextract to extract structured entities (countries, legislation, agreement types, financial amounts) from unstructured prose.

  2. Tags every cabinet decision with a policy stance and scope using qwen3, then surfaces decisions that pull in opposite directions on the same scope via qwen3-embedding retrieval + catching, for example, a permit system that re-allows products an earlier decision banned.

LLM extraction has errors, so don’t take everything at screen value and do factcheck.

Loading enriched decisions…