---
language:
- en
license:
- mit
task_categories:
- question-answering
- visual-question-answering
task_ids:
- visual-question-answering
- multiple-choice-qa
pretty_name: HippoVlog
size_categories:
- 1K<n<10K
tags:
- multimodal
- video
- memory
- reasoning
- vlogs
---

# Dataset Card for HippoVlog

## Dataset Description

- **Homepage:** [https://hippomultimodalmemory.github.io/](https://hippomultimodalmemory.github.io/)
- **Repository:** [GitHub](https://github.com/hippomultimodalmemory/hippovlog)
- **Paper:** [Link](https://arxiv.org/pdf/2504.10739)

### Dataset Summary

HippoVlog is a novel benchmark dataset designed for evaluating Multimodal Memory and Reasoning (MMR) systems. It consists of 25 long-form daily vlogs (682 minutes total) with naturalistic audiovisual content and 1,000 validated multiple-choice question-answer pairs.

### Supported Tasks and Leaderboards

The dataset supports the following tasks:
- **Multimodal Memory and Reasoning (MMR)**: The primary task involves answering questions about audiovisual content from long-form vlogs, requiring both memory and reasoning capabilities.
- **Cross-modal Understanding**: Questions test the ability to link visual and auditory information.
- **Temporal Reasoning**: Many questions require understanding events and their relationships across time.

### Languages

The dataset is in English.

## Dataset Structure

### Data Instances

Each instance in the dataset is a JSON object with the following fields:
```json
{
    "video_id": "string",          # YouTube video ID
    "question_text": "string",     # The question to be answered
    "options": {                   # Multiple choice options
        "A": "string",
        "B": "string",
        "C": "string",
        "D": "string"
    },
    "correct_answer": "string",    # The correct answer (A, B, C, or D)
    "explanation": "string",       # Explanation of the correct answer
    "category": "string"          # Question category (e.g., "audio_visual")
}
```

### Data Fields

- `video_id`: Identifier for the source video
- `question_text`: The question being asked
- `options`: Dictionary containing four possible answers labeled A through D
- `correct_answer`: The letter corresponding to the correct answer
- `explanation`: Detailed explanation of why the answer is correct
- `category`: The type of question (e.g., audio_visual, temporal, etc.)

## Dataset Creation

### Curation Rationale

HippoVlog addresses key limitations in existing multimodal understanding benchmarks by providing:
- Long-form continuous videos with naturalistic content
- Ground truth answers for objective evaluation
- Rich audio-visual integration challenges
- Temporal reasoning requirements