add eval of math_for_5th-grader

[mochisky]

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

math_for_5th-grader

Eval description

Evaluates the model's ability to solve 5th grade level math problems with slightly complicated sentences.

What makes this a useful eval?

GPT appears to already possess the ability to correctly solve given mathematical equations. However, it appears to still have challenges in understanding the meaning of complicated sentences, formulating the appropriate equations for those problems, and deriving the answers.

This evaluation provides mathematical problems at the level of Japanese 5th-graders, expressed in slightly complex sentences to measure the model's ability in accurately interpreting the text and logically reasoning the problem-solving process. Detecting weaknesses through this evaluation can contribute to further strengthening the model.

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 commits on 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": "You are a mathematician with high reading comprehension skills. You will reason before answering the following question. Your final answer will be only with numbers. No explanation needed."}, {"role": "user", "content": "What is the sum of the interior angles of a decagon?"}], "ideal": "1440"}
{"input": [{"role": "system", "content": "You are a mathematician with high reading comprehension skills. You will reason before answering the following question. Your final answer will be only with numbers. No explanation needed."}, {"role": "user", "content": "What is the least common multiple of 36, 54, and 72?"}], "ideal": "216"}
{"input": [{"role": "system", "content": "You are a mathematician with high reading comprehension skills. You will reason before answering the following question. Your final answer will be only with numbers. No explanation needed."}, {"role": "user", "content": "How many milliliters is 7.6 deciliters?"}], "ideal": "760"}
{"input": [{"role": "system", "content": "You are a mathematician with high reading comprehension skills. You will reason before answering the following question. Your final answer will be only with numbers. No explanation needed."}, {"role": "user", "content": "According to a rule, how many is the 15th number from the left when the numbers are arranged as follows: 70, 67, 64, 61, 58, ..., 7, 4, 1"}], "ideal": "28"}
{"input": [{"role": "system", "content": "You are a mathematician with high reading comprehension skills. You will reason before answering the following question. Your final answer will be only with numbers. No explanation needed."}, {"role": "user", "content": "There is beef priced at 240 yen for 80g. How much would it cost to buy 150g of this beef?"}], "ideal": "450"}
{"input": [{"role": "system", "content": "You are a mathematician with high reading comprehension skills. You will reason before answering the following question. Your final answer will be only with numbers. No explanation needed."}, {"role": "user", "content": "There have been several math tests so far, and the average score was 80 points. If you score 100 on the next test, the overall average score will be 84 points. How many tests have there been so far?"}], "ideal": "4"}
{"input": [{"role": "system", "content": "You are a mathematician with high reading comprehension skills. You will reason before answering the following question. Your final answer will be only with numbers. No explanation needed."}, {"role": "user", "content": "There is a circle with a diameter of 20cm. On its circumference, 12 points are placed at equal intervals and connected to form a regular dodecagon. What is the area of this regular dodecagon in square centimeters?"}], "ideal": "300"}
{"input": [{"role": "system", "content": "You are a mathematician with high reading comprehension skills. You will reason before answering the following question. Your final answer will be only with numbers. No explanation needed."}, {"role": "user", "content": "Mike, John and Steve had a total of 48 cards. First, Mike gave one-fifth of his cards to John. Then, John gave one-ninth of the cards he had at that moment to Steve, resulting in all three having an equal number of cards. How many cards did John have initially?"}], "ideal": "14"}
{"input": [{"role": "system", "content": "You are a mathematician with high reading comprehension skills. You will reason before answering the following question. Your final answer will be only with numbers. No explanation needed."}, {"role": "user", "content": "I bought some oranges for 20 yen each. I threw away 8 of the oranges that were rotten. I sold the rest for 45 yen each, resulting in a total profit of 2,140 yen. How many oranges did I purchase?"}], "ideal": "100"}

[usama-openai]

Thanks for the contribution. I would like to request some changes.

I would recommend asking the model to reason before answering and using the Includes evaluation method here because logical reasoning and other complex tasks are hard for the model to do zero-shot without a chance to reason through the steps. Asking the model to provide reasoning will give the model a fair chance to solve the question. You can add instructions to provide the chain of thought before answering and to provide the final output in a specific format, and then use that format to write the ideal answer. For example, the model can be asked to enclose the final answer in square brackets, and the ideal answer can be formatted like [4]. It'll give the model a proper chance to reason before answering, and the proper formatting will help in identifying the final answer using the Includes method.

Instead of writing, You will reason before answering the following question. Your final answer will be only with numbers. No explanation needed.; write something like, You will reason before answering the following question. Your final answer will be enclosed in square brackets. Only write the numeric value in brackets., and format the ideal answer as "ideal": "[4]".

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

[mochisky]

@usama-openai Thanks for your review! I've made the requested changes.

[usama-openai]

Thanks for updating the prompt. You're also supposed to be using the Includes evaluation method, but the evaluation method being used is still Match. Kindly fix this in the yaml file.

[mochisky]

@usama-openai Thanks for pointing that out. I fixed yaml file to use the Includes evaluation method. Please check it again.

[usama-openai]

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