freePruner: A Training-free Approach for Large Multimodal Model Acceleration
About
Large Multimodal Models (LMMs) have demonstrated impressive capabilities in visual-language tasks but face significant deployment challenges due to their high computational demands. While recent token reduction methods show promise for accelerating LMMs, they typically require extensive retraining or fine-tuning, making them impractical for many state-of-the-art models, especially those with proprietary training data. We propose freePruner, a training-free token reduction approach that can be directly applied to any open-source LMM without additional training. Unlike existing methods that rely heavily on token merging operations, freePruner employs a two-stage token selection strategy: (1) identifying pivotal tokens that capture high-level semantic information using our designed contribution degree metric, and (2) selecting complementary tokens that preserve essential low-level visual details through attention pattern analysis. Extensive experiments demonstrate that freePruner achieves 2x acceleration while maintaining comparable performance across mainstream visual question-answering benchmarks in the training-free setting. Moreover, freePruner is orthogonal to and can be combined with other post-training acceleration techniques, such as post-training quantization, providing a practical solution for efficient LMM deployment.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Video Question Answering | MSRVTT-QA | Accuracy59.5 | 513 | |
| Video Question Answering | ActivityNet-QA | Accuracy48.4 | 438 | |
| Video Question Answering | MSVD-QA | Accuracy71.3 | 401 | |
| Object Hallucination Evaluation | POPE | Accuracy87.7 | 259 | |
| Visual Question Answering | VQA v2 | Accuracy77.6 | 257 | |
| Multimodal Evaluation | MME | MME Score1.49e+3 | 179 | |
| Text-based Visual Question Answering | TextVQA VQAT | Accuracy60 | 71 | |
| Multimodal Evaluation | MM-Bench | Accuracy63.8 | 67 | |
| Visual Question Answering | ScienceQA SQA-I | Accuracy68.6 | 21 |