{"ID":2859724,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2510.04533","arxiv_id":"2510.04533","title":"TAG: Tangential Amplifying Guidance for Hallucination-Resistant Sampling","abstract":"Diffusion models achieve state-of-the-art image generation but often produce semantic inconsistencies, or hallucinations. Existing inference-time guidance methods rely on external signals or architectural modifications, adding computational overhead. We propose $\\mathbf{T}$angential $\\mathbf{A}$mplifying $\\mathbf{G}$uidance $\\mathbf{(TAG)}$, a training-free, architecture-agnostic, plug-and-play guidance method that operates purely on trajectory signals. TAG uses an intermediate sample as a projection basis and amplifies the tangential components of the estimated score to correct the sampling trajectory. A first-order Taylor analysis shows that this steers the state toward higher-probability regions of the data manifold, reducing inconsistencies and improving fidelity while adding negligible overhead to existing samplers. Code is available at our Project Page (https://hyeon-cho.github.io/TAG/).","short_abstract":"Diffusion models achieve state-of-the-art image generation but often produce semantic inconsistencies, or hallucinations. Existing inference-time guidance methods rely on external signals or architectural modifications, adding computational overhead. We propose $\\mathbf{T}$angential $\\mathbf{A}$mplifying $\\mathbf{G}$ui...","url_abs":"https://arxiv.org/abs/2510.04533","url_pdf":"https://arxiv.org/pdf/2510.04533v2","authors":"[\"Hyunmin Cho\",\"Donghoon Ahn\",\"Susung Hong\",\"Jee Eun Kim\",\"Seungryong Kim\",\"Kyong Hwan Jin\"]","published":"2025-10-06T06:53:29Z","proceeding":"cs.CV","tasks":"[\"cs.CV\"]","methods":"[\"Diffusion Model\"]","has_code":false}