← Dictionary / Compute-Optimal Strategy

Compute-Optimal Strategy

Appears in 1 paper

The choice of which inference-time strategy (Best-of-N vs.

As used in Paper 23 — Scaling LLM Test-Time Compute Optimally Can be More Effective than Scaling Model Parameters →

The choice of which inference-time strategy (Best-of-N vs. sequential revision, and how many attempts or rounds) maximises accuracy for a given token budget. The optimal strategy depends on problem difficulty and whether feedback from one attempt helps subsequent attempts.

Appears in papers

Paper 23 — Scaling LLM Test-Time Compute Optimally Can be More Effective than Scaling Model Parameters →