{"id":"W4416035046","doi":"10.18653/v1/2025.findings-emnlp.387","title":"AdaptMerge: Inference Time Adaptive Visual and Language-Guided Token Merging for Efficient Large Multimodal Models","year":2025,"lang":"","type":"article","venue":"","topic":"Multimodal Machine Learning Applications","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Saskatchewan","funders":"Natural Sciences and Engineering Research Council of Canada; DeepMind","keywords":"Inference; Security token; Feature (linguistics); Visualization; Pattern recognition (psychology)","routes":{"ca_aff":true,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.001004181,0.0006087465,0.0006467827,0.0004535172,0.0009161753,0.0004175185,0.001123429,0.0002562132,0.0001617097],"category_scores_gemma":[0.0004117425,0.0006127984,0.0002053241,0.001058134,0.0001684279,0.0004534516,0.001532874,0.0005330115,0.0001067728],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001670187,"about_ca_system_score_gemma":0.000452984,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001994711,"about_ca_topic_score_gemma":0.00005301962,"domain_scores_codex":[0.9957592,0.0002543677,0.0008451789,0.001599698,0.0004692638,0.001072257],"domain_scores_gemma":[0.9968268,0.001150711,0.0002874663,0.0009404488,0.0004815436,0.0003129991],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0002669504,0.001074488,0.0004523587,0.0001815985,0.0002966468,0.000006255964,0.009188128,0.5316746,0.003637238,0.3626763,0.0008757595,0.08966965],"study_design_scores_gemma":[0.002087764,0.0002689518,0.001821148,0.0001468406,0.00009154146,0.000002669895,0.0003437745,0.9923074,0.0007237989,0.001332006,0.000270487,0.0006036098],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1045907,0.0007142393,0.8854785,0.001846741,0.0002504025,0.001883898,0.00007667665,0.0003606831,0.004798179],"genre_scores_gemma":[0.8741419,0.0000214608,0.1218028,0.0005335218,0.00007591889,0.0003339236,0.0000230121,0.00003474801,0.003032668],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.7695513,"threshold_uncertainty_score":0.9996324,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0188835584601864,"score_gpt":0.3386773881967409,"score_spread":0.3197938297365545,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}