Papers with long-form video understanding
Language Repository for Long Video Understanding (2025.findings-acl)
Copied to clipboard
| Challenge: | Language-based learning models (LLMs) support long context-lengths but their effectiveness in handling long-term information gradually declines with input length. |
| Approach: | They propose a Language Repository (LangRepo) that maintains concise and structured information as an interpretable representation. |
| Outcome: | The proposed framework is evaluated on zero-shot visual question-answering benchmarks. |