Research map / Large language models / Language, safety and interpretability
ZeroUnlearn: Few-Shot Knowledge Unlearning in Large Language Models
Yujie Lin, Chengyi Yang, Zhishang Xiang, YIPING SONG, Jinsong Su
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use ZeroUnlearn instead of AlphaEdit | Llama-3.1 · MCF few-shot | efficacy (Eff.): 0.20 -> 0.00 | Improved |
| use ZeroUnlearn instead of AlphaEdit | Llama-3.2 · MCF few-shot | efficacy (Eff.): 2.60 -> 0.40 | Improved |
| use ZeroUnlearn instead of AlphaEdit | Llama-3.2 · MCF few-shot | generalization (Gen.): 11.80 -> 4.60 | Improved |
| use ZeroUnlearn instead of AlphaEdit | Llama-3.2 · MCF few-shot | specificity (Spe.): 18.36 -> 14.90 | Hurt |
| use ZeroUnlearn instead of AlphaEdit | Llama-3.2 · ZsRE few-shot | efficacy (Eff.): 29.59 -> 27.85 | Improved |
| use ZeroUnlearn instead of GA | Llama-3.2 · MCF few-shot | PPL: >1000 -> 13.06 | Improved |
| use ZeroUnlearn instead of FT | Llama-3.2 · MCF few-shot | specificity (Spe.): 0.00 -> 14.90 | Improved |
| use ZeroUnlearn instead of ROME | Llama-3.1 · MCF few-shot | efficacy (Eff.): 24.40 -> 0.00 | Improved |
| use ZeroUnlearn instead of MEMIT | Llama-3.1 · MCF few-shot | efficacy (Eff.): 9.60 -> 0.00 | Improved |
| use ZeroUnlearn-GD instead of AlphaEdit for 1000-sample unlearning | Llama-3.2 · MCF multiple unlearning | efficacy (Eff.): 1.40 -> 0.00 | Improved |
| use ZeroUnlearn-GD instead of AlphaEdit for 1000-sample unlearning | Llama-3.2 · MCF multiple unlearning | specificity (Spe.): 15.68 -> 12.41 | Hurt |
| use ZeroUnlearn-GD instead of AlphaEdit for 1000-sample unlearning | Llama-3.2 · ZsRE multiple unlearning | efficacy (Eff.): 27.66 -> 25.29 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- AlphaEdit — AlphaEdit: Null-Space Constrained Knowledge Editing for Language Models (ICLR 2025) · better on 4 of 7 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- In-Context Unlearning: Language Models as Few-Shot Unlearners (ICML 2024 · similarity 0.93)
- Large Language Model Unlearning via Embedding-Corrupted Prompts (NeurIPS 2024 · similarity 0.92)
- On Large Language Model Continual Unlearning (ICLR 2025 · similarity 0.91)
- Towards Robust and Parameter-Efficient Knowledge Unlearning for LLMs (ICLR 2025 · similarity 0.91)
- A Closer Look at Machine Unlearning for Large Language Models (ICLR 2025 · similarity 0.91)
- RWKU: Benchmarking Real-World Knowledge Unlearning for Large Language Models (NeurIPS 2024 · similarity 0.91)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
