---
license: apache-2.0
datasets:
- jeopardy
language:
- en
---

# Jeopardy Bot

Jeopardy Bot is a reasobably good and fast bot at answering jeopardy questsions. Jeopardy is a great format for Language Models because the query is typically very short and the answer is typically even shorter.

The `https://huggingface.co/datasets/jeopardy` dataset used to train this model is relatively incomplete with ~216K rows, but provides a rich evaluation set of data as https://j-archive.com/ provides daily updated clues and answers totaling ~468K clues and answers.

# Sample Responses
Here are some sampled questsions taken from recent episodes. We use Daily Doubles since those are often considered "harder".

```
Below is a Jeopardy clue paired with input providing the category of the clue. Write a concise response that best answers tbe clue given the category.

### Instruction:
In 2023 16th St. Baptist in this city marks 150 years of the congregation, a history marred by a 1963 bombing

### Input:
Take me to church

### Response:
what is Birmingham
```
Correct! Daily Double! https://j-archive.com/suggestcorrection.php?clue_id=461177

```
Below is a Jeopardy clue paired with input providing the category of the clue. Write a concise response that best answers tbe clue given the category.

### Instruction:
You can wear your Crocs as you see the gators in this national park established in 1947 at Florida's southwestern tip

### Input:
Florida Places

### Response:
what is Everglades National Park
```
Correct! Daily Double! https://j-archive.com/suggestcorrection.php?clue_id=461065


Let's Just give it the Basic Alpaca prompt with no input or even a category...
```
Below is an instruction that describes a task. Write a response that appropriately completes the request.

### Instruction:
On May 10-11, 1927 he flew from San Diego to New York City, with an overnight stop in St. Louis

### Response:
what is Charles Lindbergh
```
Correct! Daily Double! https://j-archive.com/suggestcorrection.php?clue_id=459838

```
Below is an instruction that describes a task. Write a response that appropriately completes the request.

### Instruction:
He was born March 30, 1970, in a Virginia stable

### Response:
what is George Washington
```
Womp womp womp! Let's give it the category inline...

```
Below is an instruction that describes a task. Write a response that appropriately completes the request.

### Instruction:
He was born March 30, 1970, in a Virginia stable. The Category is Sports GOATs.

### Response:
what is Michael Jordan
```
Well, even large language models aren't going to get these right all the time.

# Training
This model is SFT on LLaMa 7B. The training hyperparameters are provided in the YML config, which can be used directly with Axolotl. Training was stopped after ~2 epochs as the evaluation loss had started to increase after that. See https://wandb.ai/wing-lian/jeopardy-bot-7b