Researchers have introduced Evidence-Backed Video Question Answering (E-VQA), a new task designed to make Video Large Language Models (Video LLMs) more transparent. Current models often provide answers without clear visual justification, and existing explainability methods are limited. E-VQA requires models to output both a semantic answer and precise spatio-temporal evidence, such as temporal segments and object masks. To support this, a new benchmark called ST-Evidence has been created, along with a large-scale dataset, ST-Evidence-Instruct, to train models for fine-grained visual grounding. AI
IMPACT This research could lead to more trustworthy and interpretable video AI systems by demanding verifiable visual evidence for answers.
RANK_REASON The cluster describes a new research paper introducing a novel task and benchmark for video question answering models.
- alphaXiv
- arXiv
- CatalyzeX
- DagsHub
- E-VQA
- Gotit.pub
- Hugging Face
- Influence Flower
- ScienceCast
- ST-Evidence
- ST-Evidence-Instruct
- UniPixel
- Video LLMs
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
