CRAFT: Compression via Recursive Adaptive Fusion of Video Tokens for Vision-Language Models · 相关技术
相关技术
DINDELDAPTCross-Spatial Hybrid AttentionContextual Residual AggregationConfidence-Guided Early StoppingClinical Vision-Language ModelsCalibrated Entropy ScoreCalibrated Collective OversightCRISCOSECOMAutomatic Text ScoringAutomatic Depression DetectionAugmented Latent IntrinsicsArtificial Language LearningAmodal Mask Prediction and ROI Re-Alignment (AMPR)Advanced Persistent ThreatsAdaptive Context CompressionAdaptive Batch ScalingAction UnitsAccelerated failure time (AFT) modelsATEANNAIME 2025AILS-AHDAI-based Visually Impaired AssistanceAHATCRaFTAbstractACT