Neuroplasticity and Corruption in Model Mechanisms: A Case Study Of Indirect Object Identification
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866909523877298176 |
|---|---|
| author | Chhabra, Vishnu Kabir Zhu, Ding Khalili, Mohammad Mahdi |
| author_facet | Chhabra, Vishnu Kabir Zhu, Ding Khalili, Mohammad Mahdi |
| contents | Previous research has shown that fine-tuning language models on general tasks enhance their underlying mechanisms. However, the impact of fine-tuning on poisoned data and the resulting changes in these mechanisms are poorly understood. This study investigates the changes in a model's mechanisms during toxic fine-tuning and identifies the primary corruption mechanisms. We also analyze the changes after retraining a corrupted model on the original dataset and observe neuroplasticity behaviors, where the model relearns original mechanisms after fine-tuning the corrupted model. Our findings indicate that: (i) Underlying mechanisms are amplified across task-specific fine-tuning which can be generalized to longer epochs, (ii) Model corruption via toxic fine-tuning is localized to specific circuit components, (iii) Models exhibit neuroplasticity when retraining corrupted models on clean dataset, reforming the original model mechanisms. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_01896 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Neuroplasticity and Corruption in Model Mechanisms: A Case Study Of Indirect Object Identification Chhabra, Vishnu Kabir Zhu, Ding Khalili, Mohammad Mahdi Machine Learning Artificial Intelligence Previous research has shown that fine-tuning language models on general tasks enhance their underlying mechanisms. However, the impact of fine-tuning on poisoned data and the resulting changes in these mechanisms are poorly understood. This study investigates the changes in a model's mechanisms during toxic fine-tuning and identifies the primary corruption mechanisms. We also analyze the changes after retraining a corrupted model on the original dataset and observe neuroplasticity behaviors, where the model relearns original mechanisms after fine-tuning the corrupted model. Our findings indicate that: (i) Underlying mechanisms are amplified across task-specific fine-tuning which can be generalized to longer epochs, (ii) Model corruption via toxic fine-tuning is localized to specific circuit components, (iii) Models exhibit neuroplasticity when retraining corrupted models on clean dataset, reforming the original model mechanisms. |
| title | Neuroplasticity and Corruption in Model Mechanisms: A Case Study Of Indirect Object Identification |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2503.01896 |