Tag
Ontology
2essays ·All tags
When Rubrics Become Rewards
A fair agent benchmark is not only a scoreboard. Two-arm runs with expert criteria produce the traces that become SFT, preference, and GRPO training signal — if you design the evaluation so the verdict is trustworthy.
Memory Finds. Ontology Decides.
Semantic vault recall still returns near-misses on institutional questions. We open-sourced CQE — ClawQL’s entity definition format — and a legal Matter pack that turns escrow and non-compete into typed predicates. Here is the schema, the ingest path, the query, and the scores.