Skip to content

Context Window

Definition
AI-generated

The context window is the maximum number of tokens a language model can attend to in a single forward pass—covering system instructions, user input, retrieved documents, and (for chat models) prior turns—beyond which text must be truncated, summarized, or split across calls.

Topics
LLM and Agents
Related
Context Context Engineering In-Context Learning

Why it matters in GWAS

Long methods sections, full supplementary PDFs, or wide summary-statistic tables may exceed the window; chunking, retrieval, or column-wise processing affects what the model can “see” and therefore the reliability of automated checks or summaries.

Example usage

"We chunked the meta-analysis README into overlapping segments because the full file exceeded the API context window."

Context Context Engineering In-Context Learning Large Language Model (LLM) Prompt Prompt Engineering Retrieval-Augmented Generation (RAG) Token
Related-term network

Click any node to open that term page.

Node order. The current term is first. Next come every page linked from Related terms, in the same top-to-bottom order as that list. Then new terms appear in two-hop layers: for each of those direct neighbors (in that same order), their related terms are appended if not already listed. A third pass does the same for two-hop nodes. The walk stops when a fixed maximum number of nodes is reached.

References

  • Vaswani A, et al. (2017). Attention is all you need. NeurIPS.
← Context Engineering Contig →

Last updated (UTC · Git history)