Few-Shot Examples Have Uneven Quality and Drag Output Down
You provided 5 few-shot examples. Two are great, three are mediocre. The model averages toward the mediocre ones. Why example quality variance hurts and how to curate.
#llm-output
Articles tagged with #llm-output
Model Returns Invalid JSON Because Schema Was Described, Not Enforced
You asked for JSON matching a schema. Most calls return valid JSON, some return prose with JSON inside, some omit fields. Description vs enforcement, and how to fix at the API layer.
Model Invented Fake Citations and URLs
The model produced citations like Smith et al. 2019, journal of XYZ — and the paper does not exist. Or it linked to a URL that 404s. Why citation hallucination happens and how to stop it.
Model Replies in the Wrong Language
You prompted in English and the model answered in Chinese, or vice versa. Or it switched mid-output. Why language drift happens and how to lock the output language.
Response Got Cut Off Mid-Sentence Because max_tokens Was Too Low
The model's response ends abruptly in the middle of a sentence, a JSON object, or a code block. Almost always max_tokens. How to size it, detect truncation, and recover.
Prompt Asks for 10 Items, Model Returns 3 and Stops
You asked for 10 ideas, the model gave 3 and trailed off. Or it filled 10 slots but the last 4 are filler. Why list-length requests under-deliver and how to actually get N items.
Prompt Contains a Stale Year That Anchors the Model to Wrong Context
Your prompt template still says 2023 in 2026. Model anchors to 2023 context — old API versions, old pricing, old facts. Why date-staleness compounds and how to keep prompts evergreen.