{"ID":2840191,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2511.14751","arxiv_id":"2511.14751","title":"Co-Me: Confidence-Guided Token Merging for Visual Geometric Transformers","abstract":"We propose Confidence-Guided Token Merging (Co-Me), an acceleration mechanism for visual geometric transformers without retraining or finetuning the base model. Co-Me distilled a light-weight confidence predictor to rank tokens by uncertainty and selectively merge low-confidence ones, effectively reducing computation while maintaining spatial coverage. Compared to similarity-based merging or pruning, the confidence signal in Co-Me reliably indicates regions emphasized by the transformer, enabling substantial acceleration without degrading performance. Co-Me applies seamlessly to various multi-view and streaming visual geometric transformers, achieving speedups that scale with sequence length. When applied to VGGT and Pi3, Co-Me achieves up to 21.5x and 20.4x speedup, making visual geometric transformers practical for real-time 3D perception and reconstruction.","short_abstract":"We propose Confidence-Guided Token Merging (Co-Me), an acceleration mechanism for visual geometric transformers without retraining or finetuning the base model. Co-Me distilled a light-weight confidence predictor to rank tokens by uncertainty and selectively merge low-confidence ones, effectively reducing computation w...","url_abs":"https://arxiv.org/abs/2511.14751","url_pdf":"https://arxiv.org/pdf/2511.14751v2","authors":"[\"Yutian Chen\",\"Yuheng Qiu\",\"Ruogu Li\",\"Ali Agha\",\"Shayegan Omidshafiei\",\"Jay Patrikar\",\"Sebastian Scherer\"]","published":"2025-11-18T18:52:22Z","proceeding":"cs.CV","tasks":"[\"cs.CV\",\"cs.RO\"]","methods":"[\"Transformer\"]","has_code":false}