Maybe1407/zsre_phi_to_unlearn
TEXT GENERATIONConcurrent Unit Cost:1Model Size:1.4BQuant:BF16Context Size:2kPublished:Mar 4, 2025License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold
Maybe1407/zsre_phi_to_unlearn is a 1.4 billion parameter model, likely based on the Phi architecture, designed for unlearning specific information. This model focuses on the task of removing or mitigating previously learned associations or facts. Its primary use case involves research and applications in machine unlearning and privacy-preserving AI.
Loading preview...
Model Overview
Maybe1407/zsre_phi_to_unlearn is a compact 1.4 billion parameter model, likely derived from the Phi family of small language models. Its core purpose is to explore and demonstrate machine unlearning capabilities, specifically in the context of zero-shot relation extraction (ZSRE).
Key Capabilities
- Targeted Unlearning: Designed to remove or reduce the influence of specific data or knowledge from the model's learned parameters.
- Zero-Shot Relation Extraction Context: Implies its application in scenarios where the model needs to unlearn relations without explicit retraining on negative examples.
- Compact Size: At 1.4 billion parameters, it offers a more accessible platform for experimenting with unlearning techniques compared to larger models.
Good For
- Research in Machine Unlearning: Ideal for academics and researchers investigating methods to erase specific information from LLMs.
- Privacy-Preserving AI: Relevant for exploring techniques to comply with 'right to be forgotten' requests or mitigate data leakage.
- Understanding Model Forgetting: Useful for studying how models can be made to 'forget' certain facts or associations post-training.