{
  "context_test": "Evaluating model performance across varying context window lengths to assess attention distribution, information retrieval, and instruction adherence at scale.",
  "limitation": "This request is not a real context sweep unless the server is relaunched at different ctx sizes.",
  "next_step": "Relaunch the inference server with explicitly configured context lengths, run identical prompts through each configuration, and log latency, attention metrics, and output fidelity for comparative analysis."
}