179 questions
No questions match those filters.
← All questions
LLM FoundationsMedium· Prompt Engineering· Few-Shot Learning· Model Performance
Your prompt is format-fragile. You can fund one of: a 1...
This is one of the questions in the full AI/ML interview bank. Pro unlocks all 1789 questions; Premium includes the same bank plus the highest daily Practice limit.
See plansBigger models and instruction-tuned checkpoints primarily raise mean accuracy without necessarily reducing the spread of performance across different prompt renderings. Few-shot prompting is the most effective lever for reducing this sensitivity. However, one must account for the trade-offs: increased latency, higher GPU-second consumption, and the impact on the KV cache.