Official study guide and Watsonx docs both point to B and C. QAT (B) helps keep accuracy, and quantization (C) is mostly about saving memory and speed. Seen similar phrasing on older IBM practice sets, pretty sure these are the best two but open if someone found an edge case.
B tbh, C is also right. B nails how quantization-aware training (QAT) reduces accuracy loss, which you don't get with just post-training quantization. C covers resource efficiency and is what actually happens in most setups (smaller models, similar accuracy). A feels like a trap since 8-bit quantization doesn't always wreck performance. Unless there's some super-specific edge case, B and C are the best picks.
Yeah, B and C make the most sense here. QAT (B) helps keep accuracy closer to the original, and C is all about getting better memory usage and speed with quantization. Not 100% on edge cases but these fit best.
