Research map / Video / Video-language understanding
HERMES: temporal-coHERent long-forM understanding with Episodes and Semantics
Gueter Josmy Faure, Jia-Fong Yeh, Min-Hung Chen, Hung-Ting Su, Shang-Hong Lai, Winston H. Hsu
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 |
|---|---|---|---|
| replace MA-LMM memory bank with ECO | MA-LMM · MovieChat-1k (zero-shot, plug-in) | accuracy: 73.3 -> 76.7 | Improved |
| replace MA-LMM memory bank with ECO | MA-LMM · MovieChat-1k (zero-shot, plug-in) | inference time: 1 -> 0.569 (-43%) | Improved |
| add SeTR to MA-LMM | MA-LMM · MovieChat-1k (zero-shot, plug-in) | accuracy: 73.3 -> 77.1 | Improved |
| add SeTR to MA-LMM | MA-LMM · MovieChat-1k (zero-shot, plug-in) | inference time: +1.5% | No clear effect |
| integrate ECO into LongVA and LLaVA-OneVision | LongVA, LLaVA-OneVision · VideoMME and MovieChat-1k | latency: 30% and 35% reduction | Improved |
| add SeTR on top of ECO | LongVA, LLaVA-OneVision · VideoMME and MovieChat-1k | accuracy: +0.37% | Improved |
| add positional encoding before ECO (ablation) | HERMES · MovieChat-1k (zero-shot, global) | accuracy: 77.3 -> 78.6 | Improved |
| add SeTR (ablation) | HERMES · MovieChat-1k (zero-shot, global) | accuracy: 5% drop in accuracy when removed | Improved |
| use ECO update strategy instead of random or FIFO retention (ablation) | HERMES · MovieChat-1k (zero-shot, global) | accuracy | Improved |
| use hierarchical Q-Former instead of flat frame or video Q-Former (ablation) | HERMES · Breakfast | accuracy: 95.2% | Improved |
| use ECO instead of ToMe and MA-LMM memory (ablation) | HERMES · MovieChat-1k (zero-shot, global) | accuracy: 78.6% vs 13.8% above ToMe, 5.3% above MA-LMM | Improved |
| increase input frames from 40 to 1000 (ablation) | HERMES · MovieChat-1k (zero-shot, global) | accuracy: 77.6% -> 76.7% | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- MA-LMM — MA-LMM: Memory-Augmented Large Multimodal Model for Long-Term Video Understanding (CVPR 2024) · better on 3 of 3 shared results
- MovieChat — MovieChat: From Dense Token to Sparse Memory for Long Video Understanding (CVPR 2024) · better on 3 of 3 shared results
- Movies2Scenes — Movies2Scenes: Using Movie Metadata To Learn Scene Representation (CVPR 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- One Token per Highly Selective Frame: Towards Extreme Compression for Long Video Understanding (NeurIPS 2025 · similarity 0.89)
- SEAL: Semantic Attention Learning for Long Video Representation (CVPR 2025 · similarity 0.88)
- Streaming Long Video Understanding with Large Language Models (NeurIPS 2024 · similarity 0.88)
- Leveraging Textual Memory and Key Frame Reasoning for Full Video Understanding Using Off-the-Shelf LLMs and VLMs (Student Abstract) (AAAI 2025 · similarity 0.88)
- LongVideoBench: A Benchmark for Long-context Interleaved Video-Language Understanding (NeurIPS 2024 · similarity 0.88)
- Vgent: Graph-based Retrieval-Reasoning-Augmented Generation For Long Video Understanding (NeurIPS 2025 · similarity 0.88)
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.
