{"work":{"id":"4152c935-00bb-47e6-a371-51ba28c2e58d","openalex_id":null,"doi":null,"arxiv_id":"2410.02355","raw_key":null,"title":"AlphaEdit: Null-Space Constrained Knowledge Editing for Language Models","authors":null,"authors_text":"URLhttps: //arxiv","year":2024,"venue":"cs.CL","abstract":"Large language models (LLMs) often exhibit hallucinations due to incorrect or outdated knowledge. Hence, model editing methods have emerged to enable targeted knowledge updates. To achieve this, a prevailing paradigm is the locating-then-editing approach, which first locates influential parameters and then edits them by introducing a perturbation. While effective, current studies have demonstrated that this perturbation inevitably disrupt the originally preserved knowledge within LLMs, especially in sequential editing scenarios. To address this, we introduce AlphaEdit, a novel solution that projects perturbation onto the null space of the preserved knowledge before applying it to the parameters. We theoretically prove that this projection ensures the output of post-edited LLMs remains unchanged when queried about the preserved knowledge, thereby mitigating the issue of disruption. Extensive experiments on various LLMs, including LLaMA3, GPT2-XL, and GPT-J, show that AlphaEdit boosts the performance of most locating-then-editing methods by an average of 36.7% with a single line of additional code for projection solely. Our code is available at: https://github.com/jianghoucheng/AlphaEdit.","external_url":"https://arxiv.org/abs/2410.02355","cited_by_count":null,"metadata_source":"pith","metadata_fetched_at":"2026-07-05T16:21:15.970863+00:00","pith_arxiv_id":"2410.02355","created_at":"2026-05-10T12:05:22.113016+00:00","updated_at":"2026-07-05T16:21:15.970863+00:00","title_quality_ok":true,"display_title":"Alphaedit: Null-space constrained knowledge editing for language models","render_title":"Alphaedit: Null-space constrained knowledge editing for language models"},"hub":{"state":{"work_id":"4152c935-00bb-47e6-a371-51ba28c2e58d","tier":"hub","tier_reason":"10+ Pith inbound or 1,000+ external citations","pith_inbound_count":25,"external_cited_by_count":null,"distinct_field_count":5,"first_pith_cited_at":"2025-07-04T17:21:46+00:00","last_pith_cited_at":"2026-06-25T15:18:31+00:00","author_build_status":"not_needed","summary_status":"needed","contexts_status":"needed","graph_status":"needed","ask_index_status":"not_needed","reader_status":"not_needed","recognition_status":"not_needed","updated_at":"2026-08-22T21:09:41.146341+00:00","tier_text":"hub"},"tier":"hub","role_counts":[{"context_role":"background","n":4},{"context_role":"baseline","n":1},{"context_role":"method","n":1}],"polarity_counts":[{"context_polarity":"background","n":4},{"context_polarity":"baseline","n":1},{"context_polarity":"use_method","n":1}],"runs":{},"summary":{},"graph":{},"authors":[]}}