Cheng Jin, Zhongtian Ma, Xiao Wang, Han Qi, Xiaoyu Wen, Xiang Wang, Youbang Sun, Xia Hu · 2026 · arXiv
DOI: 10.48550/arXiv.2610.25791 · open-access PDF
Grandes modelos de linguagem adquirem amplo conhecimento da web-scale corpora durante o pré-treinamento, incluindo conhecimento que poderia ser perigoso se usado mal. Remover esse conhecimento do modelo em si sem degradar a capacidade geral é desejável, mas difícil, pois está profundamente entrelaçado com a capacidade geral dentro dos mesmos parâmetros. Os métodos de alinhamento de segurança existentes e de não aprendizagem normalmente ajustam esses parâmetros compartilhados, tendendo a suprimir a expressão de conhecimento perigoso em vez de removê-lo.
Machine translation (Meta M2M100); the English original follows.
Large language models acquire broad knowledge from web-scale corpora during pretraining, including knowledge that could be hazardous if misused. Removing such knowledge from the model itself without degrading general capability is desirable but difficult, as it is deeply entangled with general capability within the same parameters. Existing safety alignment and unlearning methods typically adjust these shared parameters, tending to suppress the expression of hazardous knowledge rather than remove it. As a result, the knowledge remains in the model and is readily recoverable by subsequent fine-
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.25791, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange