[evals] Add 3D Globe Movement Eval

[thair116]

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 be 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 it 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. Starting April 10, the minimum eval count is 15 samples, we hope this makes it easier to create and contribute evals.

Also, please note that we're using Git LFS for storing the JSON files, so please make sure that you move the JSON file to Git LFS before submitting a PR. Details on how to use Git LFS are available here.

Eval details 📑

Eval name

3D Globe Movement

Eval description

This eval tests an LLMs ability to understand 3D movement through space, in particular movement on or through planet Earth. Each example provides a starting point and a path consisting of one or two movements, and the expected answer is a state/province or ocean.

Similar to the evals from #462 and #1060, this eval shows how difficult movement is for LLMs to understand, and builds upon those by showing how the problem is seemingly magnified by 3D movement and/or by requesting a region as answer rather than numerical positions.

Testing on gpt-3.5-turbo, accuracy ranges from ~0.24 to ~0.31

What makes this a useful eval?

This eval demonstrates that a long series of steps is not necessary in order to create a path that GPT is unable to follow, and that a simple trip to the planet's core and back again, with a slight offset in any direction, will often get it lost.

Whereas it can often handle "travel 15 degrees East", hiding that actual travel behind a 2-step 3D path significantly hurts the model's performance.

Interestingly, this eval seems to demonstrate the statistical rather than reasoning nature of GPT, as regardless of starting point or path, it shows a bias towards answering with oceans, especially the Pacific Ocean.

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).

Your eval should be:

• 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.
• Contains failures where a human can do the task, but either GPT-4 or GPT-3.5-Turbo could not.
• 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.
• Include at least 15 high-quality examples.

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

• Check that your data is in evals/registry/data/{name}
• Check that your YAML is registered at evals/registry/evals/{name}.yaml
• 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).

• 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.

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

Limited availability acknowledgment

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 the 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.

• 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 be granted.

Submit eval

• I have filled out all required fields of this form
• I have used Git LFS for the Eval JSON data
• (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": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of San Francisco and travel directly downwards to the center of the Earth, then straight up, back to the surface, at a 15 degree offset to the East"}], "ideal": "Colorado"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of San Francisco and travel due East 15 degrees"}], "ideal": "Colorado"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of San Francisco and travel due East 0.1 degrees"}], "ideal": "California"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of San Francisco and travel due East 1 meter"}], "ideal": "California"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of San Francisco and travel 0 meter"}], "ideal": "California"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of San Francisco and make one complete revolution around the planet"}], "ideal": "California"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of San Francisco and travel directly downwards to the center of the Earth, then straight up, back to the surface, at a 1 degree offset to the East"}], "ideal": "California"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of San Francisco and travel directly downwards to the center of the Earth, then straight up, back to the surface, at a 0.1 degree offset to the East"}], "ideal": "California"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of San Francisco and travel directly downwards to the center of the Earth, then straight up, back to the surface, at a 0.01 degree offset to the East"}], "ideal": "California"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of San Francisco and travel directly downwards to the center of the Earth, then straight up, back to the surface, at a 1 degree offset to the West"}], "ideal": "Pacific Ocean"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of San Francisco and travel directly downwards to the center of the Earth, then straight up, back to the surface, at a 90 degree offset to the East"}], "ideal": "Atlantic Ocean"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of San Francisco and travel directly downwards to the center of the Earth, then straight up, back to the surface, at a 180 degree offset to the East"}], "ideal": "Ahal"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of San Francisco and travel directly downwards to the center of the Earth, then straight up, back to the surface, at a 1 degree offset to the North"}], "ideal": "California"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of Kansas City and travel directly downwards to the center of the Earth, then straight up, back to the surface, at a 15 degree offset to the East"}], "ideal": "West Virginia"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of Kansas City and travel directly downwards to the center of the Earth, then straight up, back to the surface, at a 15 degree offset to the West"}], "ideal": "Utah"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of New York City and travel directly downwards to the center of the Earth, then straight up, back to the surface, at a 1 degree offset to the East"}], "ideal": "Atlantic Ocean"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of New York City and travel due East 10 degrees"}], "ideal": "Atlantic Ocean"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of New York City and travel due East 500 kilometers"}], "ideal": "Atlantic Ocean"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of New York City and travel directly downwards to the center of the Earth, then straight up, back to the surface, at a 1 degree offset to the West"}], "ideal": "New Jersey"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of New York City and travel directly downwards to the center of the Earth, then straight up, back to the surface, at a 10 degree offset to the West"}], "ideal": "Ohio"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of New York City and travel due West 10 degrees"}], "ideal": "Ohio"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of New York City and travel due West 15 degrees"}], "ideal": "Illinois"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of New York City and travel due West 25 degrees"}], "ideal": "Nebraska"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of New York City and travel due West 30 degrees"}], "ideal": "Colorado"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of New York City and travel due West 35 degrees"}], "ideal": "Colorado"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of Oklahoma City and travel directly downwards to the center of the Earth, then straight up, back to the surface, at a 0.1 degree offset to the East"}], "ideal": "Oklahoma"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of Oklahoma City and travel directly downwards to the center of the Earth, then straight up, back to the surface along the same path that was traveled downards."}], "ideal": "Oklahoma"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of Oklahoma City and travel directly downwards to the center of the Earth, then continue onwards to the other side of the planet"}], "ideal": "Xinjiang"}
{"input": [{"role": "system", "content": "Pretend you are 3DPathGPT, a model that can accurately track 3D paths on, around, and through planet Earth based on plain language descriptions. For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe]. Start at the center of Oklahoma City and travel directly downwards through to the other side of the Earth"}], "ideal": "Xinjiang"}

[usama-openai]

Thanks for the contribution. I would like to request a change.

The prompt in the samples is ambiguous. First, it asks the model, For each path, determine the state, province, or ocean of the final destination. Provide just that as an answer and no other text, and then it asks the model, Try reasoning through the 3D path one step at a time. which is contradictory. The evaluation method being used is Includes, which is able to detect the ideal answer in long completion. I would suggest you update the prompt as follows:

For each path, determine the state, province, or ocean of the final destination. You may assume the object traveling the path encounters no resistance, such as if it were a neutrino. Try reasoning through the 3D path one step at a time, and at the end, provide the final answer enclosed in square brackets like [Europe].

The ideal answer can be formatted like [Europe]. It'll give the model a proper chance to reason before answering, and the proper formatting will help in identifying the final answer properly using the Includes method.

We would love to review the PR again after the suggested changes.

[thair116]

Thank you for the feedback. With that change made the accuracy on gpt3 has improved slightly from ~0.24-0.31 to ~0.28-0.41 (3 runs each). I'll be happy to make any other improvements you can think of

[thair116]

I forgot to add the brackets around the answer in the previous commit. Adding that too brought the accuracy to ~0.31-0.34

[usama-openai]

This PR looks in good shape now. I'm approving this PR.

[thair116]

thanks, @usama-openai. I saw that the build was failing so I rebased to main. Please let me know if there's anything else I need to do to get the build to pass

[usama-openai]

@thair116 You don't need to do anything else.

[thair116]

Looks like all checks have passed now. Can we get this merged today? 🙂

[usama-openai]

You should see GPT-4 API access enabled in your account in the next few days.