[Fundamentals of Machine Learning and Neural Networks]
What are the main advantages of instructed large language models over traditional, small language models (< 300M parameters)? (Pick the 2 correct responses)
Instructed large language models (LLMs), such as those supported by NVIDIA's NeMo framework, have significant advantages over smaller, traditional models:
Option D: LLMs often have cheaper computational costs during inference for certain tasks because they can generalize across multiple tasks without requiring task-specific retraining, unlike smaller models that may need separate models per task.
Option E: A single generic LLM can perform multiple tasks (e.g., text generation, classification, translation) due to its broad pre-training, unlike smaller models that are typically task-specific.
Option A is incorrect, as LLMs require large amounts of data, often labeled or curated, for pre-training. Option B is false, as LLMs typically have higher latency and lower throughput due to their size. Option C is misleading, as LLMs are often less interpretable than smaller models.
NVIDIA NeMo Documentation: https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/stable/nlp/intro.html
Brown, T., et al. (2020). 'Language Models are Few-Shot Learners.'
Colby
8 months agoRaina
8 months agoZona
8 months agoJacki
9 months agoFrederic
9 months agoAlaine
9 months agoGladis
9 months agoChau
9 months agoJess
10 months agoAlecia
10 months agoLisbeth
10 months agoGlenna
10 months agoClarence
10 months agoGilma
11 months agoJoana
1 year agoKina
1 year agoCheryl
1 year agoWilliam
1 year agoHannah
1 year agoCordelia
1 year agoLorean
1 year agoTanesha
1 year agoMike
1 year agoVincent
1 year agoHerminia
1 year agoDeane
1 year ago