Glossary · Term

capability buffer

← all terms

Definition

Plain language

The way a safety scaffold absorbs damage when you switch to a weaker model — like a seat belt mattering more in worse conditions.

As stated in the literature

The observation in TraceFix experiments that a verified coordination protocol absorbs roughly half the task-completion degradation incurred when downgrading the underlying LLM to a weaker model.

Why it matters: It means good agent scaffolding doesn't just help frontier models — it cushions the impact of using cheaper or smaller models too.

For example, swapping in a weaker LLM under the TraceFix protocol cut task completion by only about half as much as swapping it in without the protocol.

Mentioned in 1 episode

  1. 034
    Catching Multi-Agent Deadlocks Before Deployment With a 40-Year-Old Tool

Related terms