TTF: Temporal Token Fusion for Efficient Video-Language Model

arXiv:2605.07355v1 Announce Type: cross Abstract: Video-language models (VLMs) face rapid inference costs as visual token counts scale with video length. For example, 32 frames at 448{imes}448 resolution already yield >8,000 visual tokens in Qwen3-VL, making LLM prefill the dominant throughput bottleneck. Existing methods often rely on global similarity or attention-guided compression, incurring offsets to their gains. We propose extbf{Temporal Token Fusion (TTF)}, a training-free, plug-and-play pre-LLM token compression framework that exploits structured temporal redundancy in video. TTF automatically selects an anchor frame, then for each subsequent frame, performs a local window similarity search (e.g.,3imes 3), fusing tokens that exceed a threshold. The compressed sequence maintains positional consistency across both prefill and decoding through coordinate realignment, enabling seamless integration with existing VLM pipelines. On Qwen3-VL-8B with threshold t=0.70, TTF removes about 67% of visual tokens while retaining 99.5% of the baseline accuracy and introducing only {approx}0.16,GFLOPs of matching overhead. Overall, TTF offers a practical, efficient solution for video understanding. The code is available at href{https://github.com/Cominder/ttf}{https://github.com/Cominder/ttf}

Source: arXiv cs.AI | 2026-05-11