GPT-3 (Brown et al., 2020) performed tasks from a few examples placed in the prompt, with no weight updates. The prompt is just text: a few solved input-output pairs, then the new input with its answer left blank for the model to fill in.
Implement build_few_shot_prompt(examples, query, sep), which formats each example as input => output, appends query =>, and joins everything with sep.
Build a list of formatted strings, add the query as the last item, then join with the separator.
No results yet
Run your code to check it against the examples, or submit to run every test.