Human body movement

[tescao]

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

Human body movement (sample size=10, will contribute more if accepted)

Eval description

An eval to asses the model's understanding of how a human body moves, e.g. after performing an exercise starting from a known side, which side does the human end up with?

What makes this a useful eval?

The eval test the model's physical reasoning abilities as well as good knowledge of a human body, which is necessary for being safe and helpful to users, and could be used in fitness applications, as an aid for people with hearing impairment, generating dance sequences, etc.

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:

• [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.
• [ ] 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 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.

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}.yaml
• [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": "Answer the question based on the following description of a movement. Assume average human body ability."}, {"role": "user", "content": "\nDescription: A woman is standing on her hands and feet, facing the floor and reaching her hips towards the ceiling, she lifts her left leg towards the ceiling, then lowers it and puts it in between her hands. She inhales and twists her torso pressing her chest against her inner thigh and reaches her arm to the ceiling.\nQuestion: Which arm does she reach up?\nAnswer with a single word: 'left' or 'right'."}], "ideal": "left"}
{"input": [{"role": "system", "content": "Answer the question based on the following description of a movement. Assume average human body ability."}, {"role": "user", "content": "\nDescription: A woman is standing on her left leg, she lifts her other leg and presses it into her chest with both arms. She then grabs the heel of her lifted leg with her hand and stretches it all the way to the side, still holding the heel. She remains upright with her chest opened.\nQuestion: Which hand does she use to hold her leg?\nAnswer with a single word: 'left' or 'right'."}], "ideal": "right"}
{"input": [{"role": "system", "content": "Answer the question based on the following description of a movement. Assume average human body ability."}, {"role": "user", "content": "\nDescription: A woman is holding a reversed plank position, holding her body in a straight line and facing up. She lifts her left leg off the ground then lifts her hand off the ground and transitions into a side plank.\nQuestion: Which hand does she have to lift off the ground to transition into side plank?\nAnswer with a single word: 'left' or 'right'."}], "ideal": "left"}
{"input": [{"role": "system", "content": "Answer the question based on the following description of a movement. Assume average human body ability."}, {"role": "user", "content": "\nDescription: A man is in a plank position, holding his body in a straight line and facing the floor. Bracing his core, he lifts his left leg off the floor until it is level with his body, then lifts his opposite arm until it is level with his body, then puts them back on the floor. This is one repetition and he does eight such repetitions, starting with his left leg and alternating sides. This is one set. He takes a few seconds to rest then does the same set again, starting with his right leg this time.\nQuestion: Which arm does he lift at the last repetition of the last set?\nAnswer with a single word: 'left' or 'right'."}], "ideal": "right "}
{"input": [{"role": "system", "content": "Answer the question based on the following description of a movement. Assume average human body ability."}, {"role": "user", "content": "\nDescription: A woman stands straight, she takes a big step back with her left foot, turning toes of that foot out slightly. She lengthens her torso, opening her chest and reaching her side towards her front leg until she can grab her leg with her arm. She reaches the other arm to the sky and holds, lengthening her body and opening her chest more.\nQuestion: Which arm did she reach up?\nAnswer with a single word: 'left' or 'right'."}], "ideal": "left"}

[Ein-Tim]

Please fill in the INSERT_EVAL_HERE field and give this PR a meaningful title.

[tescao]

@andrew-openai that's awesome, thanks for the update! I'd be happy to add some more shortly, 10 seems kinda unreliable? Could you please direct me to the instructions on how to move the file to git-lfs though? I can only see them regarding fetching evals for running..

[andrew-openai]

Once you have git-lfs installed (git lfs install), it should automatically move them to lfs on push.

[tescao]

@andrew-openai hi, gently pinging to see if you've had a chance to look at the new commit. I fixed lfs and added some more data.

[tescao]

@usama-openai hey thanks for noticing, fixed the spaces!

[tescao]

@usama-openai thanks for merging! Can I still get the GPT-4 access?