Research map / Video / Video-language understanding
EventFlash: Towards Efficient MLLMs for Event-Based Vision
Shaoyu Liu, Jianing Li, guanghui zhao, Yunjian Zhang, Wen Jiang, Ming Li, Xiangyang Ji
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add spatiotemporal token sparsification (spatial and temporal) (ablation) | EventFlash-Zero · EventMind | throughput (tokens/s): 2.3 -> 28.5 (+12.4x) | Improved |
| add spatiotemporal token sparsification (ablation) | EventFlash-Zero · EventMind | GDC / FGQA / HAQA / MCQA: GDC 45.3 -> 46.8; FGQA 60.4 -> 61.1; HAQA 85.0 -> 84.9; MCQA 58.2 -> 60.0 | No clear effect |
| add temporal sparsification (ATWA) only (ablation) | EventFlash-Zero · EventMind | throughput (tokens/s): 2.3 -> 14.0 (+6.1x) | Improved |
| add spatial sparsification (SDGA) only (ablation) | EventFlash-Zero · EventMind | throughput (tokens/s): 2.3 -> 5.3 (+2.3x) | Improved |
| add spatial sparsification only (ablation) | EventFlash-Zero · EventMind | GDC / FGQA / HAQA / MCQA: GDC 45.3 -> 46.3; FGQA 60.4 -> 61.2; HAQA 85.0 -> 85.1; MCQA 58.2 -> 59.6 | No clear effect |
| add temporal sparsification only (ablation) | EventFlash-Zero · EventMind | HAQA: 85.0 -> 83.8 | Hurt |
| increase aggregation interval from 10ms to 20ms (ablation) | EventFlash · EventMind | GDC: 46.8 -> 43.2 | Hurt |
| increase aggregation interval from 10ms to 20ms (ablation) | EventFlash · EventMind | throughput (tokens/s): 28.5 -> 52.6 | Improved |
| increase aggregation interval from 10ms to 30ms (ablation) | EventFlash · EventMind | MCQA: 84.9 -> 61.8 | Hurt |
| decrease aggregation interval from 10ms to 5ms (ablation) | EventFlash · EventMind | throughput (tokens/s): 28.5 -> 15.8 | Hurt |
| decrease aggregation interval from 10ms to 5ms (ablation) | EventFlash · EventMind | GDC / FGQA: GDC 46.8 -> 47.1; FGQA 61.1 -> 61.8 | No clear effect |
| increase density attenuation factor alpha from 0.1 to 0.6 (ablation) | EventFlash · EventMind | MCQA: 84.9 -> 83.2 | Hurt |
Closest papers by meaning
- EventGPT: Event Stream Understanding with Multimodal Large Language Models (CVPR 2025 · similarity 0.92)
- Talk2Event: Grounded Understanding of Dynamic Scenes from Event Cameras (NeurIPS 2025 · similarity 0.89)
- Learning to See through Illumination Extremes with Event Streaming in Multimodal Large Language Models (CVPR 2026 · similarity 0.89)
- RE-VLM: Event-Augmented Vision-Language Model for Scene Understanding (CVPR 2026 · similarity 0.88)
- EventZoom: A Progressive Approach to Event-Based Data Augmentation for Enhanced Neuromorphic Vision (AAAI 2025 · similarity 0.87)
- ReVisionLLM: Recursive Vision-Language Model for Temporal Grounding in Hour-Long Videos (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
