{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/unlearning-isn-t-deletion-investigating","title":"Unlearning Isn't Deletion: Investigating Reversibility of Machine Unlearning in LLMs","arxiv_id":"2505.16831","date":"2025-05-22","proceeding":null,"authors":["Xiaoyu Xu","Xiang Yue","Yang Liu","Qingqing Ye","Haibo Hu","Minxin Du"],"abstract":"Unlearning in large language models (LLMs) is intended to remove the influence of specific data, yet current evaluations rely heavily on token-level metrics such as accuracy and perplexity. We show that these metrics can be misleading: models often appear to forget, but their original behavior can be rapidly restored with minimal fine-tuning, revealing that unlearning may obscure information rather than erase it. To diagnose this phenomenon, we introduce a representation-level evaluation framework using PCA-based similarity and shift, centered kernel alignment, and Fisher information. Applying this toolkit across six unlearning methods, three domains (text, code, math), and two open-source LLMs, we uncover a critical distinction between reversible and irreversible forgetting. In reversible cases, models suffer token-level collapse yet retain latent features; in irreversible cases, deeper representational damage occurs. We further provide a theoretical account linking shallow weight perturbations near output layers to misleading unlearning signals, and show that reversibility is modulated by task type and hyperparameters. Our findings reveal a fundamental gap in current evaluation practices and establish a new diagnostic foundation for trustworthy unlearning in LLMs. We provide a unified toolkit for analyzing LLM representation changes under unlearning and relearning: https://github.com/XiaoyuXU1/Representational_Analysis_Tools.git.","url_abs":"https://arxiv.org/abs/2505.16831v1","url_pdf":"https://arxiv.org/pdf/2505.16831v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"unlearning-isn-t-deletion-investigating","repo_url":"https://github.com/xiaoyuxu1/representational_analysis_tools","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"diagnostic","task_name":"Diagnostic"},{"task_slug":"machine-unlearning","task_name":"Machine Unlearning"},{"task_slug":"math","task_name":"Math"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2505.16831","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2505.16831"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/xiaoyuxu1/representational_analysis_tools","reach":{"status":"ok","spdx":"MIT"}}],"summary":{"unverified":4},"by_repo_kind":{"official":{"samples":4,"ran":0,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"e20ef443e94ec77a","entry":"run_cka_analysis","repo":"xiaoyuxu1/representational_analysis_tools","repo_kind":"official","path":"representational_analysis/src/representational_toolkit/cka_analysis.py","file_url":"https://github.com/xiaoyuxu1/representational_analysis_tools/blob/HEAD/representational_analysis/src/representational_toolkit/cka_analysis.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"e20ef443e94ec77a"}},{"code_sha256_prefix":"a1e37080f7e8169b","entry":"run_fim_analysis","repo":"xiaoyuxu1/representational_analysis_tools","repo_kind":"official","path":"representational_analysis/src/representational_toolkit/fisher_analysis.py","file_url":"https://github.com/xiaoyuxu1/representational_analysis_tools/blob/HEAD/representational_analysis/src/representational_toolkit/fisher_analysis.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"a1e37080f7e8169b"}},{"code_sha256_prefix":"8b3a56ecc95d9ddf","entry":"run_pca_shift","repo":"xiaoyuxu1/representational_analysis_tools","repo_kind":"official","path":"representational_analysis/src/representational_toolkit/pca_shift_analysis.py","file_url":"https://github.com/xiaoyuxu1/representational_analysis_tools/blob/HEAD/representational_analysis/src/representational_toolkit/pca_shift_analysis.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"8b3a56ecc95d9ddf"}},{"code_sha256_prefix":"328e0583d4914fbd","entry":"run_pca_similarity","repo":"xiaoyuxu1/representational_analysis_tools","repo_kind":"official","path":"representational_analysis/src/representational_toolkit/pca_sim_analysis.py","file_url":"https://github.com/xiaoyuxu1/representational_analysis_tools/blob/HEAD/representational_analysis/src/representational_toolkit/pca_sim_analysis.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"328e0583d4914fbd"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}