hillclimb

An optimization method that repeatedly applies small changes and keeps those that raise an evaluation score.

1 articleLast mentioned

Hill climbing, often written hillclimb or hillclimbing, is a search and optimization method that repeatedly makes a small change to the current state, keeps the change if an evaluation score improves, and reverts it otherwise. It is a classic technique in mathematical optimization and artificial intelligence search. As a heuristic, it can get stuck in a local optimum rather than reaching the best overall solution, and convergence is not guaranteed.

In AI engineering, the term describes the practice of iteratively adjusting prompts, settings, or code and measuring each change against an eval score. To avoid overfitting to the evaluation data, the data is commonly split into train and test sets, and a change is kept only when scores improve on both.

This entry is based on AIPOST articles and widely known facts. If something is wrong, please send us a correction request.

Articles covering this entry

  • Prompt Audits and Cache Reuse Cut Claude Costs Up to 72.9%, Anthropic SaysAI in Practice ·

    The /claude-api hillclimb command builds a defense into its loop.

    Read article

Found a mistake or missing an entry?

Wiki entries are based on AIPOST articles and widely known facts. If an entry is wrong, send us a correction request and we will check and fix it. Names from new articles are added to the wiki after the article is published.