-
Taken for Granted? It's time we reconsider the Instruction Tuning Loss! 馃憠
Uncovering the benefits of Weighted Instruction Tuning (WIT)
-
Towards Understanding Multi-Task Learning (Generalization) of LLMs via Detecting and Exploring Task-Specific Neurons
-
Layer by Layer: Uncovering Where Multi-Task Learning Happens in Instruction-Tuned Large Language Models
-
Transformer-Based Language Models