KiToke: Distribution-Aware Token Compression for Video Large Language Models

arXiv:2604.03414v2 Announce Type: replace Abstract: Video Large Language Models (Video LLMs) achieve strong performance on video understanding tasks but suffer from high inference costs due to the large number of visual tokens. We study training-free, query-agnostic video token compression, where heuristic importance rules can over-represent large redundant visual groups and miss less frequent visual patterns, particularly under aggressive token budgets. We propose KiToke, a distribution-aware compression framework that treats visual tokens as samples from the video's spatiotemporal token distribution. KiToke uses global embedding density as a proxy for representational redundancy and inverse-density sampling to balance token allocation across visual patterns of different frequencies. It further aggregates unselected tokens into retained anchors within detected temporal intervals to reduce cross-transition mixing. Across evaluated backbones, KiToke achieves the highest average accuracy among compared training-free methods and remains robust down to 1% retention, where most evaluated baselines underperform uniform random token sampling. On LLaVA-OneVision, it preserves 98.5% and 89.2% of the uncompressed model's four-benchmark mean accuracy at 10% and 1% token retention, with 6.8x and 12.5x prefill speedups, respectively, including compression overhead.

Related

• EvoCut: Multi-Layer Evolution-Aware Visual Token Compression for Efficient Large Vision-Language Models
• O-MARC: Omni Memory-Augmented Compression Distillation for Efficient Video Understanding
• Persistent Object Narratives for Token-Efficient Video Language Models
• GeoTrace: Geometry-Aware Trajectory Token Compression for Video Large Language Models
• Are We Using the Right Benchmark: An Evaluation Framework for Visual Token Compression Methods

Source: arXiv cs.CV | 2026-10-15