Generalization of Transformers with In-Context Learning: An Empirical Study

Generalization of Transformers with In-Context Learning: An Empirical Study [ICLR 2025]

Xingxuan Zhang, Haoran Wang, Jiansheng Li, Yuan Xue, Shikai Guan, Renzhe Xu, Hao Zou, Han Yu, Peng Cui

📝 Paper | 🤗 Huggingface | 📚 Slide

The code repository contains code for all of our main experiments.

• embed-level function fitting
• token-level function fitting
• translation
• api-calling

As each task can be split into preprocess, train and evaluation clearly, you can read the comments of the corresponding source code to get usage.

For submodules, you can use git submodule update --init to get them.

Thanks for those repos:

• https://github.com/hiyouga/LLaMA-Factory/
• https://github.com/dtsip/in-context-learning/
• https://github.com/Blealtan/efficient-kan/
• https://github.com/THUNLP-MT/StableToolBench/