In-context learning

What In-context learning means

In-context learning is a language-model concept about context selection and limits. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output. The important distinction is between a model setting or interface label and the behavior a user can observe. The same concept can have different effects across model families, tokenizers, prompts, datasets, and serving interfaces. A useful explanation states what the concept changes, which inputs it depends on, and what evidence would support a claim about it.

How it appears in practice

Consider an engineering assistant that receives a repository question, selects context, and returns a proposed change. In-context learning may influence one stage of that interaction, while the final result also depends on model version, instructions, tools, stopping rules, and review. Preserve those conditions when comparing runs. If the concept is internal, use an external task or evaluation to test its practical effect instead of assuming that a score or setting has a simple meaning.

Example and measurement

A team could evaluate In-context learning with ordinary requests and boundary cases. Define expected behavior before running the comparison, then record completion quality, test results, latency, input and output tokens, retries, and correction effort. For example, a code workflow should include real repository context and a verification step rather than judging a fluent standalone answer. Inspect changed cases individually because aggregate averages can hide regressions in a critical task.

Limitations

In-context learning is not a substitute for testing, source review, or authorization. Results can shift after a model update, prompt change, data change, or provider change. The concept may explain one part of a request without proving that the answer is factual, safe, or useful. Document the exact implementation and evaluation conditions when using In-context learning to make quality, cost, or routing decisions.