2truths1lie (27% accuracy)

[hackgoofer]

Thank you for contributing an eval! ♥️

🚨 Please make sure your PR follows these guidelines, failure to follow the guidelines below will result in the PR being closed automatically. Note that even if the criteria are met, that does not guarantee the PR will be merged nor GPT-4 access granted. 🚨

PLEASE READ THIS:

In order for a PR to be merged, it must fail on GPT-4. We are aware that right now, users do not have access, so you will not be able to tell if the eval fails or not. Please run your eval with GPT-3.5-Turbo, but keep in mind as we run the eval, if GPT-4 gets higher than 90% on the eval, we will likely reject since GPT-4 is already capable of completing the task.

We plan to roll out a way for users submitting evals to see the eval performance on GPT-4 soon. Stay tuned! Until then, you will not be able to see the eval performance on GPT-4. We encourage partial PR's with ~5-10 example that we can then run the evals on and share the results with you so you know how your eval does with GPT-4 before writing all 100 examples.

Eval details 📑

Eval name

2Truths1Lie

Eval description

I took a list of historical events (from wikipedia both ancient and modern), and wrote a script to generate 2 "facts" and 1 "lie" with those events. The lie has the wrong date always for the historical event.

What makes this a useful eval?

GPT4 is trained with the massive corpus online, and it should be able to detail important historical events with little to no trouble. I also noticed that gpt4 is inconsistent in its response. See below:

Criteria for a good eval ✅

Below are some of the criteria we look for in a good eval. In general, we are seeking cases where the model does not do a good job despite being capable of generating a good response (note that there are some things large language models cannot do, so those would not make good evals). Humans can easily look it up and figure out the right answer to those questions.

Your eval should be:

• [X] Thematically consistent: The eval should be thematically consistent. We'd like to see a number of prompts all demonstrating some particular failure mode. For example, we can create an eval on cases where the model fails to reason about the physical world.
• [X] Contains failures where a human can do the task, but either GPT-4 or GPT-3.5-Turbo could not.
• [X] Includes good signal around what is the right behavior. This means either a correct answer for Basic evals or the Fact Model-graded eval, or an exhaustive rubric for evaluating answers for the Criteria Model-graded eval.
• [X] Include at least 100 high quality examples (it is okay to only contribute 5-10 meaningful examples and have us test them with GPT-4 before adding all 100)

If there is anything else that makes your eval worth including, please document it below.

Unique eval value

Insert what makes your eval high quality that was not mentioned above. (Not required)

Eval structure 🏗️

Your eval should

• [X] Check that your data is in evals/registry/data/{name}
• [X] Check that your yaml is registered at evals/registry/evals/{name}.jsonl
• [X] Ensure you have the right to use the data you submit via this eval

(For now, we will only be approving evals that use one of the existing eval classes. You may still write custom eval classes for your own cases, and we may consider merging them in the future.)

Final checklist 👀

Submission agreement

By contributing to Evals, you are agreeing to make your evaluation logic and data under the same MIT license as this repository. You must have adequate rights to upload any data used in an Eval. OpenAI reserves the right to use this data in future service improvements to our product. Contributions to OpenAI Evals will be subject to our usual Usage Policies (https://platform.openai.com/docs/usage-policies).

• [X] I agree that my submission will be made available under an MIT license and complies with OpenAI's usage policies.

Email address validation

If your submission is accepted, we will be granting GPT-4 access to a limited number of contributors. Access will be given to the email address associated with the merged pull request.

• [X] I acknowledge that GPT-4 access will only be granted, if applicable, to the email address used for my merged pull request.

Limited availability acknowledgement

We know that you might be excited to contribute to OpenAI's mission, help improve our models, and gain access to GPT-4. However, due to the requirements mentioned above and high volume of submissions, we will not be able to accept all submissions and thus not grant everyone who opens a PR GPT-4 access. We know this is disappointing, but we hope to set the right expectation before you open this PR.

• [X] I understand that opening a PR, even if it meets the requirements above, does not guarantee the PR will be merged nor GPT-4 access granted.

Submit eval

• [X] I have filled out all required fields in the evals PR form
• [X] (Ignore if not submitting code) I have run pip install pre-commit; pre-commit install and have verified that black, isort, and autoflake are running when I commit and push

Failure to fill out all required fields will result in the PR being closed.

Eval JSON data

Since we are using Git LFS, we are asking eval submitters to add in as many Eval Samples (at least 5) from their contribution here:

View evals in JSON

Eval

{"input": [{"role": "system", "content": "Tell me which one of the following 3 is a lie? Only use numeric index as the answer."}, {"role": "user", "content": "0: The Algerian Civil War ends on February 8, 2002; 1: Golden Jubilee of Elizabeth II on February 6, 2003; 2: Oldest known depiction of the Staff God, the oldest image of a god to be found in the America in 2250 BCE; "}], "ideal": "1"}
{"input": [{"role": "system", "content": "Tell me which one of the following 3 is a lie? Only use numeric index as the answer."}, {"role": "user", "content": "0: An earthquake strikes Gujarat, India, on Republic Day, resulting in more than 20,000 deaths on January 26, 2001; 1: Crash of American Airlines Flight 587 on November 12, 2001; 2: Mozilla Firefox is released on September 23, 2003; "}], "ideal": "2"}
{"input": [{"role": "system", "content": "Tell me which one of the following 3 is a lie? Only use numeric index as the answer."}, {"role": "user", "content": "0: Kot Diji phase of the Indus Valley Civilisation begins in 2800 BCE; 1: The Euro enters circulation on January 1, 2002; 2: SpaceX is founded by Elon Musk on March 14, 2003; "}], "ideal": "2"}
{"input": [{"role": "system", "content": "Tell me which one of the following 3 is a lie? Only use numeric index as the answer."}, {"role": "user", "content": "0: Alphabetic writing emerge in 1801 BCE; 1: Eleven members of the royal family of Nepal, including the king and queen, are killed by Crown Prince Dipendra, who wounds himself and dies three days later on June 1, 2001; 2: The Hallstatt culture begin in 1200 BCE; "}], "ideal": "0"}
{"input": [{"role": "system", "content": "Tell me which one of the following 3 is a lie? Only use numeric index as the answer."}, {"role": "user", "content": "0: Domestication of the horse in the Yamnaya culture in 3000 BCE; 1: Caral-Supe civilization begins in Peru in 3201 BCE; 2: September 11 attack on September 11, 2001; "}], "ideal": "1"}