Futurex-Online

# Submission Guidelines — Weekly Prediction Challenge

📄 Technical Report: [https://arxiv.org/pdf/2508.11987](https://arxiv.org/pdf/2508.11987)
🌐 LeaderBoard: [https://futurex-ai.github.io/](https://futurex-ai.github.io/)

We run a weekly real-time prediction challenge. This repo **always** contains latest events.

## 1. Weekly Rules

* A new set of tasks is released every week.
* **This week's tasks cover events with an end time between `2025-12-03, 24:00 (UTC+8)` and `2025-12-09, 24:00 (UTC+8)`.**
* **Option A:** If you submit your API, we will automatically test your model.
* **Option B:** If you run locally, you must download the tasks, make predictions, and email us your results **before Wednesday 24:00 (UTC+8)** each week.
* Late submissions will **not** be counted.
* The leaderboard for this round will be updated by **`2025-12-11, 24:00 (UTC+8, Beijing Time)`**.

**⚠️ Next submission deadline: <span style="color:red">2025-12-03, 24:00 (UTC+8, Beijing Time)</span>**

---

## 2. How to Participate?
Two ways:

### 2.1 Option A — Submit Your API

Send an email to **[[e-mail]](mailto:[e-mail])** with your model’s API (OpenAI-compatible endpoint + key).

**Email subject format:**

```
FutureX Challenge Entry — [Your Model Name]
```

**Email body should include:**

* Model details (base model, parameters, safety settings)
* Any required headers or rate limits

---

### 2.2 Option B — Run Locally & Submit Predictions

1. Download the latest tasks from our dataset:
[https://huggingface.co/datasets/futurex-ai/Futurex-Online](https://huggingface.co/datasets/futurex-ai/Futurex-Online)

2. Run your model on the tasks.

3. Email your predictions to **[[e-mail]](mailto:[e-mail])** before the deadline.

**Email subject format:**

```
FutureX Challenge Submission — [Your Model Name][Date]
```

**Submission requirements:**

* File format: **JSON** (or JSONL)
* Include dataset commit/hash you used
* Required fields: `id`, `prediction`

**Example (table view):**

| id | prediction |
| ---------- | ---------- |
| event\_001 | Yes |
| event\_002 | No |
| event\_003 | Maybe |

**Example (JSONL format):**

```json
{"id": "event_001", "prediction": "Yes"}
{"id": "event_002", "prediction": "No"}
{"id": "event_003", "prediction": "Maybe"}
```

---

## 3. Leaderboard and Privacy

* Once your model has been tested on **200+ events**, it will appear on the **overall leaderboard**.
* Weekly results will be ranked and shown on a dynamic leaderboard.
* If your model is **not** in the top 10, you may choose to keep it private or make it public.

We will keep your API keys and submission files confidential, and only use them for evaluation.

# 提交指南——每周预测挑战赛

📄 技术报告：[https://arxiv.org/pdf/2508.11987](https://arxiv.org/pdf/2508.11987)
🌐 排行榜：[https://futurex-ai.github.io/](https://futurex-ai.github.io/)

我们举办每周实时预测挑战赛。本仓库始终收录最新赛事动态。

## 1. 每周赛事规则

* 每周都会发布全新的任务集。
* **本周任务覆盖的事件截止时间为`2025-12-03, 24:00 (UTC+8)`至`2025-12-09, 24:00 (UTC+8)`。**
* **方案A：提交你的API接口，我们将自动对您的模型进行测试。**
* **方案B：本地运行模式：您需下载任务数据集、完成预测，并于每周三24:00（UTC+8）前将结果发送至我们的邮箱。**
* 逾期提交将不予计分。
* 本轮赛事的排行榜将在**`2025-12-11, 24:00 (UTC+8, 北京时间)`**更新完毕。

**⚠️ 下一轮提交截止日期：<span style="color:red">2025-12-03, 24:00 (UTC+8, 北京时间)</span>**

---

## 2. 参与方式

共两种参与途径：

### 2.1 方案A — 提交模型API接口

请发送邮件至**[[e-mail]](mailto:[e-mail])**，邮件中需包含您的模型API信息（兼容OpenAI格式的端点地址与密钥）。

**邮件主题格式：**

FutureX Challenge Entry — [您的模型名称]

**邮件正文需包含：**
* 模型详情（基座模型、参数量、安全配置）
* 所需的请求头或速率限制规则

---

### 2.2 方案B — 本地运行并提交预测结果

1. 从我们的数据集仓库下载最新任务：
[https://huggingface.co/datasets/futurex-ai/Futurex-Online](https://huggingface.co/datasets/futurex-ai/Futurex-Online)
2. 在本地使用您的模型完成任务推理。
3. 在提交截止日期前，将预测结果发送至**[[e-mail]](mailto:[e-mail])**。

**邮件主题格式：**

FutureX Challenge Submission — [您的模型名称][提交日期]

**提交要求：**
* 文件格式：**JSON**（或JSONL）
* 需包含所使用的数据集提交哈希（commit hash）
* 必填字段：`id`、`prediction`

**示例（表格视图）：**

| id | prediction |
| ---------- | ---------- |
| event_001 | Yes |
| event_002 | No |
| event_003 | Maybe |

**示例（JSONL格式）：**

json
{"id": "event_001", "prediction": "Yes"}
{"id": "event_002", "prediction": "No"}
{"id": "event_003", "prediction": "Maybe"}

---

## 3. 排行榜与隐私政策

* 当您的模型完成**200+个事件**的测试后，即可登上**总排行榜**。
* 每周赛事结果将按排名展示在动态更新的排行榜中。
* 若您的模型未进入Top10，您可选择将其设为私有或公开。

我们将严格保密您的API密钥与提交文件，仅用于模型评估用途。