EXTRACTION · EDITORIAL CLASSIFICATION · NOT A BENCHMARK
DO I NEED THE BIG MODEL FOR…
Moderate user-generated content
Routed-model jobMAYBE — EARN THE UPGRADE
“Nuance and asymmetric harm make blind cheap-model routing risky.”
WHY THIS VERDICTSTART WITH
START WITH
THE WORK.
- 01
The work is constrained enough to evaluate cheaply.
- 02
Several plausible interpretations can remain live.
- 03
A plausible wrong answer could create material cost or harm.
RECOMMENDED SETUPMid-tier first; route low-confidence cases upward
NON-NEGOTIABLE
Use policy-specific evals, appeals, and human escalation for severe actions.
REASONING3/5
AMBIGUITY4/5
ERROR COST5/5
CONTEXT2/5