SignalTech/AI
Claude was faster and cheaper. We took the scenic route anyway
High-priority alert based on source significance and keyword relevance.
Context
What happened: A study ran five small local models on the same questions, checked each cited source with Claude Sonnet 5.5, and used a fixed rule to pick answers.
Why it matters: The authors state the revised rule passed with a new set of models, showing it carries over to other models rather than to new questions. They also note Sonnet was better at making the first call, while their approach helps afterward because saved checks allow rule changes without new model calls.
What changed
A study ran five small local models on the same questions, checked each cited source with Claude Sonnet 5.5, and used a fixed rule to pick answers.
Why it matters
The authors state the revised rule passed with a new set of models, showing it carries over to other models rather than to new questions. They also note Sonnet was better at making the first call, while their approach helps afterward because saved checks allow rule changes without new model calls.