Instruction-follwing

Reader -> Generator RAG

RLHF (Reinformcement Learning Human Feedback)

Open Access

Self-Instruction

Document + KG

with KG

with Document

Self-attention 관련

Layernorm + initialization

multimodal

성능 향상

경량화

domain 적용

성능 향상

Distributed learning

Knowledge-Intensive Task

LM with External Knowledge

Transformer 구조 수정

Encoder-Decoder BART

Decoder-Only GPT-1

Encoder-Only BERT

Effficient Training

Benchmark

NLP