File size: 1,313 Bytes
f352454 0639664 288f7f0 0639664 288f7f0 0639664 f352454 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 |
---
license: mit
---
# README
## Introduction
This repository is designed for the AgentRecBench project, which aims to provide a comprehensive benchmark for evaluating the performance of recommendation agents. The code for this project can be found at [AgentSocietyChallenge](https://github.com/tsinghua-fib-lab/AgentSocietyChallenge/tree/final).
## Directory Structure
### `task/`
This folder contains three major types of recommendation tasks:
- **Classic Tasks**: These tasks focus on traditional recommendation scenarios where the goal is to provide accurate recommendations based on historical user-item interactions.
- **Evolving Interest Tasks**: These tasks are designed to capture the dynamic nature of user preferences over time, requiring agents to adapt to changing interests.
- **Cold-Start Tasks**: These tasks address the challenge of making recommendations for new users or items with limited interaction history.
### dataset
The following three folders contain the datasets for each task, processed and formatted for retrieval by the agent:
- **process_data_all**: Contains datasets for both classic and cold-start tasks.
- **process_data_long**: Contains datasets for long-term evolving interest tasks.
- **process_data_short**: Contains datasets for short-term evolving interest tasks. |