How code helps different tasks? A decompositional lens on LLM post-training
A decompositional lens is introduced for studying the effects of instruction-tuned models on question answering, mathematics, and code generation in LLM post-training and highlights how the value of code data in post training depends on which categories are combined for which model and task.
Zheng Yu, Yi-Wei Li, Yi-Shen Chen et al.
· 0 citations