JoaoBoer/tofu_Llama-3.1-8B-Instruct_forget01_PDU

TEXT GENERATIONPricing:Input $0.2 / Cached $0.028 / Output $0.32Concurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Oct 1, 2026License:llama3.1Architecture:Transformer Featherless Exclusive Cold

JoaoBoer/tofu_Llama-3.1-8B-Instruct_forget01_PDU is an 8-billion parameter instruction-tuned Llama-3.1 model developed by JoaoBoer, specifically unlearned on the TOFU 'forget01' split using the PDU (Primal-Dual Unlearning) method. This model is designed as a weight-unlearning baseline within the Speculative-Decoding-Unlearning project, focusing on evaluating and demonstrating machine unlearning capabilities. It excels at demonstrating the effectiveness of unlearning specific data while aiming to retain general model utility, making it suitable for research into data privacy and model editing.

Loading preview...

Overview

JoaoBoer/tofu_Llama-3.1-8B-Instruct_forget01_PDU is an 8-billion parameter instruction-tuned Llama-3.1 model that has undergone a specific unlearning process. Developed by JoaoBoer, this model was unlearned on the TOFU forget01 split using the Primal-Dual Unlearning (PDU) method, as part of the open-unlearning framework. It serves as a weight-unlearning baseline within the Speculative-Decoding-Unlearning project, demonstrating how models can be modified to 'forget' specific training data.

Key Capabilities

  • Demonstrates Machine Unlearning: Specifically engineered to showcase the effectiveness of the PDU method in removing knowledge of the forget01 split from the TOFU dataset.
  • Research Baseline: Provides a foundational model for further research into unlearning techniques and their impact on model behavior and privacy.
  • Evaluated for Forgetting: Includes detailed metrics such as exact_memorization (0.0397), extraction_strength (0.0291), and forget_quality (0.0000), indicating successful unlearning of specific information.
  • Retains Utility: Despite unlearning, the model aims to maintain general utility, with a model_utility score of 0.6518.

Good For

  • Machine Unlearning Research: Ideal for researchers studying data privacy, model editing, and the efficacy of unlearning algorithms.
  • Benchmarking Unlearning Methods: Can be used as a comparative baseline for developing and testing new unlearning techniques.
  • Understanding Model Behavior: Provides insights into how unlearning affects model performance and knowledge retention.

Finetunes
1 models