Papers by Mathew Huerta-Enochian
Instruction Fine-Tuning: Does Prompt Loss Matter? (2024.emnlp-main)
Copied to clipboard
| Challenge: | Recent research in language modeling has made huge advances in training instruction-following agents. |
| Approach: | They analyze the effects of various prompt loss token weights for supervised instruction fine-tuning. |
| Outcome: | The proposed model outperforms models fine-tuned on short-completion data on multiple-choice and short-generation benchmarks. |