Research map / Video / Video-language understanding
Long-range Modeling and Processing of Multimodal Event Sequences
Jichu Li, Yilun Zhong, Zhiting Li, Feng Zhou, Quyu Kong
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 adaptive temporal-similarity sequence compression (ablation) | MM-TPP · DanmakuTPP | RMSE: 5.5551 -> 5.2987 | Improved |
| add adaptive temporal-similarity sequence compression (ablation) | MM-TPP · DanmakuTPP | ACC (%): 25.87 -> 27.62 | Improved |
| use adaptive compression instead of random event drop p=25% (ablation) | MM-TPP · DanmakuTPP | RMSE: 5.3792 -> 5.2987 | Improved |
| use adaptive compression instead of random event drop p=50% (ablation) | MM-TPP · DanmakuTPP | ACC (%): 24.79 -> 27.62 | Improved |
| add image input to text-only MM-TPP (ablation) | MM-TPP · DanmakuTPP | RMSE: 5.4654 -> 5.2987 | Improved |
| add image input to text-only MM-TPP (ablation) | MM-TPP · DanmakuTPP | ACC (%): 23.64 -> 27.62 | Improved |
| add image input to text-only MM-TPP (ablation) | MM-TPP · TAXI-PRO | ACC (%): 76.70 -> 77.56 | Improved |
| scale backbone from Qwen2.5-VL-3B to 7B (ablation) | MM-TPP · DanmakuTPP | RMSE: 5.2987 -> 5.0533 | Improved |
| scale backbone from Qwen2.5-VL-3B to 7B (ablation) | MM-TPP · DanmakuTPP | ACC (%): 27.62 -> 26.98 | Hurt |
| scale backbone from Qwen2.5-VL-3B to 7B (ablation) | MM-TPP · TAXI-PRO | ACC (%): 77.56 -> 76.16 | Hurt |
| increase max token limit from 2048 to 4096 (ablation) | MM-TPP · DanmakuTPP | RMSE and ACC | Improved |
| use similarity threshold 0.2 instead of 0.05 or 0.5 (ablation) | MM-TPP · DanmakuTPP | RMSE and ACC | Improved |
Closest papers by meaning
- DanmakuTPPBench: A Multi-modal Benchmark for Temporal Point Process Modeling and Understanding (NeurIPS 2025 · similarity 0.89)
- EasyTPP: Towards Open Benchmarking Temporal Point Processes (ICLR 2024 · similarity 0.86)
- Seq2Time: Sequential Knowledge Transfer for Video LLM Temporal Grounding (CVPR 2025 · similarity 0.85)
- Interacting Diffusion Processes for Event Sequence Forecasting (ICML 2024 · similarity 0.85)
- SO-Bench: A Structural Output Evaluation of Multimodal LLM (CVPR 2026 · similarity 0.85)
- Human-inspired Episodic Memory for Infinite Context LLMs (ICLR 2025 · similarity 0.85)
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.
