rm sample freeform, some docs #603

andrew-openai · 2023-04-07T01:06:29Z

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

[Insert Eval name here]

Eval description

[Insert a short description of what your eval does here]

What makes this a useful eval?

[Insert why this eval is worth including and any additional context]

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 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

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 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.

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

I have filled out all required fields in the evals PR form
(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

INSERT_EVAL_HERE

jwang47 · 2023-04-07T19:04:31Z

README.md

@@ -1,18 +1,28 @@
 # Evals

-Evals is a framework for evaluating OpenAI models and an open-source registry of benchmarks.
+Evals is a framework for evaluating large language models (LLMs) and LLM Programs. It also includes an open-source registry of challenging benchmark evals.


We should define what LLM Programs are, or get rid of it since we don't use this terminology elsewhere. It's a bit confusing since CompletionFn is essentially LLM program AFAICT.

How about "Evals is a framework for evaluating large language models (LLMs) and other prompt completion programs"

I decided to go with Evals is a framework for evaluating large language models (LLMs) and other prompt completion functions

jwang47 · 2023-04-07T19:05:14Z

README.md

+- measure the quality of completions provided by LLMs or LLM programs, and
+- compare performance across different datasets.
+
+(April 7 update): We now support evaluating the performance of any end-to-end application logic via the [Completion Function Protocol](docs/completion-fns.md).


performance of any LLM program instead? (assuming we go with LLM programs)

How about
"We now support evaluating the performance of any prompt completion logic, like LLMs, prompt chains or tool-using agents, via the Completion Function Protocol."

jwang47 · 2023-04-07T19:07:40Z

docs/completion-fn-protocol.md

+ def get_completions(self) -> list[str]:
+ pass
+```
+The `get_completions` method returns a list of string completions. Each element should be considered a unique completion (in most cases this will be a list of length 1).


Currently at the level of the CompletionFn interface, there's no way to request multiple completions. Should we add it to the interface or leave it up to the subclasses?

I think we should leave it to the subclasses, maybe revise later.

jwang47 · 2023-04-07T19:11:21Z

docs/completion-fn-protocol.md

+We take a `prompt` representing a single sample from an eval. These prompts can be represented as either a text string or a list of messages in [OpenAI Chat format](https://platform.openai.com/docs/guides/chat/introduction). To work with the existing evals, Completion Function implementations would need to handle both types of inputs, but we provide helper functionality to convert Chat formatted messages into a text string if that is the preferred input for your program:
+```python
+from evals.prompt.base import CompletionPrompt
+
+# chat_prompt: list[dict[str, str]] -> text_prompt: str
+text_prompt = CompletionPrompt(chat_prompt).to_formatted_prompt()


I was a bit surprised by this block in terms of the ordering. Should we instead given an example implementation of a CompletionFn + CompletionResult that uses this?

jwang47 · 2023-04-07T19:12:31Z

docs/completion-fns.md

+In [run-evals.md](run-evals.md), we learned how to make calls to `oaieval` to run an eval against a completion function. Completion Functions are generalizations of model completions, where a "completion" is some text output that would be our answer to the prompt. For example, if "Who played the girl elf in the hobbit?" is our prompt, the correct completion is "Evangeline Lilly". While we can just test a model directly to see if it generates "Evangeline Lilly", we can imagine doing numerous other operations under the hood to improve our ability to answer this question, like giving the model access to a browser to look up the answer before responding. Making it easy to implement this kind of under-the-hood operators before responding is the motivation behind building Completion Functions.
+
+## How to implement completion functions
+A completion function needs to implement some interfaces that make it usable within Evals. At it's core, it is just standardizing inputs to be a text string or Chat conversation, and the output to be a list of text strings. Implementing this interface will allow you to run your logic against any eval in Evals.


At it's core -> At its core

jwang47 · 2023-04-07T19:13:01Z

docs/completion-fns.md

+In [run-evals.md](run-evals.md), we learned how to make calls to `oaieval` to run an eval against a completion function. Completion Functions are generalizations of model completions, where a "completion" is some text output that would be our answer to the prompt. For example, if "Who played the girl elf in the hobbit?" is our prompt, the correct completion is "Evangeline Lilly". While we can just test a model directly to see if it generates "Evangeline Lilly", we can imagine doing numerous other operations under the hood to improve our ability to answer this question, like giving the model access to a browser to look up the answer before responding. Making it easy to implement this kind of under-the-hood operators before responding is the motivation behind building Completion Functions.
+
+## How to implement completion functions
+A completion function needs to implement some interfaces that make it usable within Evals. At it's core, it is just standardizing inputs to be a text string or Chat conversation, and the output to be a list of text strings. Implementing this interface will allow you to run your logic against any eval in Evals.


Should we link to the Chat API docs?

jwang47 · 2023-04-07T19:14:59Z

docs/completion-fns.md

+
+The exact interfaces needed are described in detail in [completion-fn-protocol.md](completion-fn-protocol.md)
+
+We also include some example implementations inside `evals/completion_fns`. For example, using [`LangChainLLMCompletionFn`](../evals/completion_fns/langchain_llm.py), we can run evals against any base LLM available as `langchain/llms:BaseLLM` subclasses:


Wording seems a bit inaccurate. Should we instead say this?

"We can run evals against any registered CompletionFn, and these LangChainLLMCompletionFn instances in the registry can use any base LLM."

jwang47 · 2023-04-07T19:15:57Z

docs/completion-fns.md

+```
+
+## Registering Completion Functions
+Once you have written a completion function, we need to make the class visible to the `oaieval` CLI. Similar to how we register our evals, we also register Completion Functions inside `evals/registry/completion_fns` as `yaml` files. Here is the registration for our langchain LLM eval:


our langchain LLM eval -> our LangChain LLM completion function

jwang47 · 2023-04-07T19:17:02Z

docs/completion-fns.md

+my_project
+├── my_completion_fn.py
+└── my_completion_fn.yaml


I think the files would have to be under my_project/completion_fns, but should test it.

jwang47 · 2023-04-07T19:19:19Z

docs/run-evals.md


-In this example, `gpt-3.5-turbo` is the model to evaluate, and `test-match` is the eval to run. The valid model names are those which you have access to via the API. The valid eval names are specified in the YAML files under `evals/registry/evals`, and their corresponding implementations can be found in `evals/elsuite`.
+`gpt-3.5-turbo` is an implementation of the [`CompletionFn` protocol](completion-fn-protocol.md). Any implementation of the `CompletionFn` protocol can be run against `oaieval`. By default, we support calling `oaieval` with any model availableon the OpenAI API or with CompletionFunctions available in [`evals/registry/completion_fns`](../evals/registry/completion_fns/). 


This is a bit confusing since gpt-3.5-turbo is an API model, but also refers to instance of an OpenAIChatCompletionFn. Maybe instead say something along the lines of "gpt-3.5-turbo is an OpenAI model which we dynamically instantiate as a OpenAIChatCompletionFn(model=gpt-3.5-turbo)"

# 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 [Insert Eval name here] ### Eval description [Insert a short description of what your eval does here] ### What makes this a useful eval? [Insert why this eval is worth including and any additional context] ## 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 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 - [ ] 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 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. - [ ] 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 - [ ] I have filled out all required fields in the evals PR form - [ ] (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: <details> <summary>View evals in JSON</summary> ### Eval ```jsonl INSERT_EVAL_HERE ``` </details>

andrew-openai · 2023-04-07T20:55:43Z

I believe all comments are addressed - we can always reopen new PR if we want to fix more docs

just-nAnO · 2023-06-08T09:25:14Z

better move to a dedicated thread. Tks.

andrew-openai and others added 4 commits April 6, 2023 18:05

rm sample freeform, some docs

a240840

Update README.md

04b2455

Merge branch 'evals_refactor' into andrew/cleanup

6f062a3

add docs for completion functions

0b8997d

andrew-openai requested a review from jwang47 April 7, 2023 19:03

andrew-openai marked this pull request as ready for review April 7, 2023 19:03

Update completion-fn-protocol.md

0a15570

jwang47 reviewed Apr 7, 2023

View reviewed changes

andrew-openai and others added 3 commits April 7, 2023 13:10

address comments

a3bcb05

readme

5eb02c3

andrew-openai merged commit 4401cce into evals_refactor Apr 7, 2023

andrew-openai deleted the andrew/cleanup branch April 7, 2023 20:55

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

rm sample freeform, some docs #603

rm sample freeform, some docs #603

andrew-openai commented Apr 7, 2023

jwang47 Apr 7, 2023

andrew-openai Apr 7, 2023

andrew-openai Apr 7, 2023

jwang47 Apr 7, 2023

andrew-openai Apr 7, 2023

jwang47 Apr 7, 2023

andrew-openai Apr 7, 2023

jwang47 Apr 7, 2023

jwang47 Apr 7, 2023

jwang47 Apr 7, 2023

jwang47 Apr 7, 2023

jwang47 Apr 7, 2023

jwang47 Apr 7, 2023

jwang47 Apr 7, 2023

andrew-openai commented Apr 7, 2023

just-nAnO commented Jun 8, 2023


		The exact interfaces needed are described in detail in [completion-fn-protocol.md](completion-fn-protocol.md)

		We also include some example implementations inside `evals/completion_fns`. For example, using [`LangChainLLMCompletionFn`](../evals/completion_fns/langchain_llm.py), we can run evals against any base LLM available as `langchain/llms:BaseLLM` subclasses:


		In this example, `gpt-3.5-turbo` is the model to evaluate, and `test-match` is the eval to run. The valid model names are those which you have access to via the API. The valid eval names are specified in the YAML files under `evals/registry/evals`, and their corresponding implementations can be found in `evals/elsuite`.
		`gpt-3.5-turbo` is an implementation of the [`CompletionFn` protocol](completion-fn-protocol.md). Any implementation of the `CompletionFn` protocol can be run against `oaieval`. By default, we support calling `oaieval` with any model availableon the OpenAI API or with CompletionFunctions available in [`evals/registry/completion_fns`](../evals/registry/completion_fns/).

rm sample freeform, some docs #603

rm sample freeform, some docs #603

Conversation

andrew-openai commented Apr 7, 2023

Thank you for contributing an eval! ♥️

Eval details 📑

Eval name

Eval description

What makes this a useful eval?

Criteria for a good eval ✅

Unique eval value

Eval structure 🏗️

Final checklist 👀

Submission agreement

Email address validation

Limited availability acknowledgement

Submit eval

Eval JSON data

Eval

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

andrew-openai commented Apr 7, 2023

just-nAnO commented Jun 8, 2023