Exploring the Robustness of In-Context Learning with Noisy Labels
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866909186213806080 |
|---|---|
| author | Cheng, Chen Yu, Xinzhi Wen, Haodong Sun, Jingsong Yue, Guanzhang Zhang, Yihao Wei, Zeming |
| author_facet | Cheng, Chen Yu, Xinzhi Wen, Haodong Sun, Jingsong Yue, Guanzhang Zhang, Yihao Wei, Zeming |
| contents | Recently, the mysterious In-Context Learning (ICL) ability exhibited by Transformer architectures, especially in large language models (LLMs), has sparked significant research interest. However, the resilience of Transformers' in-context learning capabilities in the presence of noisy samples, prevalent in both training corpora and prompt demonstrations, remains underexplored. In this paper, inspired by prior research that studies ICL ability using simple function classes, we take a closer look at this problem by investigating the robustness of Transformers against noisy labels. Specifically, we first conduct a thorough evaluation and analysis of the robustness of Transformers against noisy labels during in-context learning and show that they exhibit notable resilience against diverse types of noise in demonstration labels. Furthermore, we delve deeper into this problem by exploring whether introducing noise into the training set, akin to a form of data augmentation, enhances such robustness during inference, and find that such noise can indeed improve the robustness of ICL. Overall, our fruitful analysis and findings provide a comprehensive understanding of the resilience of Transformer models against label noises during ICL and provide valuable insights into the research on Transformers in natural language processing. Our code is available at https://github.com/InezYu0928/in-context-learning. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2404_18191 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Exploring the Robustness of In-Context Learning with Noisy Labels Cheng, Chen Yu, Xinzhi Wen, Haodong Sun, Jingsong Yue, Guanzhang Zhang, Yihao Wei, Zeming Computation and Language Artificial Intelligence Cryptography and Security Machine Learning Optimization and Control Recently, the mysterious In-Context Learning (ICL) ability exhibited by Transformer architectures, especially in large language models (LLMs), has sparked significant research interest. However, the resilience of Transformers' in-context learning capabilities in the presence of noisy samples, prevalent in both training corpora and prompt demonstrations, remains underexplored. In this paper, inspired by prior research that studies ICL ability using simple function classes, we take a closer look at this problem by investigating the robustness of Transformers against noisy labels. Specifically, we first conduct a thorough evaluation and analysis of the robustness of Transformers against noisy labels during in-context learning and show that they exhibit notable resilience against diverse types of noise in demonstration labels. Furthermore, we delve deeper into this problem by exploring whether introducing noise into the training set, akin to a form of data augmentation, enhances such robustness during inference, and find that such noise can indeed improve the robustness of ICL. Overall, our fruitful analysis and findings provide a comprehensive understanding of the resilience of Transformer models against label noises during ICL and provide valuable insights into the research on Transformers in natural language processing. Our code is available at https://github.com/InezYu0928/in-context-learning. |
| title | Exploring the Robustness of In-Context Learning with Noisy Labels |
| topic | Computation and Language Artificial Intelligence Cryptography and Security Machine Learning Optimization and Control |
| url | https://arxiv.org/abs/2404.18191 |