Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
[evals] Basic world knowledge in Mongolian language(s) (~10% accuracy) (
openai#338) # 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 Basic world concepts and knowledge in Mongolian language ### Eval description Tests model's ability to answers simple questions in Mongolian language and its latin variant ### What makes this a useful eval? This eval contains several trivial questions regarding basic knowledge of the world, however they're written in Mongolian. GPT-3.5-Turbo shows ability to form answers in Mongolian language but rarely are its answers correct. Sometimes it shows it understands the concept of a question but fails to answer factually. Translated examples: - "How many legs do humans have?" - "Zero" - "Is fire hot or cold?" - "Cold" - "Which country borders Mongolia to the south?" - "Russia". In worse cases its answers are unrelated, nonsensical or even offensive. This eval also contains prompts in "Latin version" of Mongolian language. This version is not official and is used mostly by Mongolians on social media. The language model when prompted the same question in official Cyrillic and unofficial Latin version in the same session generally gives the same answers in respective version. ## 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. - [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: <details> <summary>View evals in JSON</summary> ### Eval ```jsonl {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Монгол улсын нийслэл юу вэ?"}], "ideal": "Улаанбаатар"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Mongol ulsiin niislel yu ve?"}], "ideal": "Ulaanbaatar"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Англи улсын нийслэл юу вэ?"}], "ideal": "Лондон"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Angli ulsiin niislel yu ve?"}], "ideal": "London"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Монголын түүхэн дэх хамгийн алдартай хүн хэн бэ?"}], "ideal": "Чингис"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Mongoliin tuuhen deh hamgiin aldartai hun hen be?"}], "ideal": ["Chinggis", "Genghis"]} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Дэлхийн зургаа дахь том элсэл цөл юу вэ?"}], "ideal": "Говь"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Delhiin zurgaa dahi hamgiin tom tsul yu ve?"}], "ideal": "Gobi"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Хүн хэдэн хөлтэй вэ?"}], "ideal": "Хоёр"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Hun heden hultei ve?"}], "ideal": "Hoyor"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Хүн хэдэн гартай вэ?"}], "ideal": "Хоёр"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Hun heden gartai ve?"}], "ideal": "Hoyor"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Дөрөв зургаа хоёрын хооронд ямар тоо байдаг вэ?"}], "ideal": "Тав"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Duruv zurgaa hoyrin hoorond yamar too baidag ve?"}], "ideal": "Tav"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Цас ямар өнгөтэй вэ?"}], "ideal": "Цагаан"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Tsas yamar ungutei ve?"}], "ideal": "Tsagaan"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Алим нь жимс үү ногоо юу?"}], "ideal": "Жимс"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Alim jims uu nogoo yu?"}], "ideal": "Jims"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Муур хэдэн хөлтэй вэ?"}], "ideal": "Дөрөв"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Muur heden hultei ve?"}], "ideal": ["Duruv","Dorov"]} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Монгол улс хойд хилээрээ ямар улстай хиллэдэг вэ?"}], "ideal": "Орос"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Mongol uls hoid hileeree yamar ulstai hilledeg ve?"}], "ideal": "Oros"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Монгол улс урд хилээрээ ямар улстай хиллэдэг вэ?"}], "ideal": "Хятад"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Mongol uls urd hileeree yamar ulstai hilledeg ve?"}], "ideal": "Hyatad"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Нэг цагийг жар хуваасны нэгийг юу гэдэг вэ?"}], "ideal": "Минут"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Neg tsagiig jar huvaasnii negiig yu gdg ve?"}], "ideal": "Minut"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Гал халуун уу хүйтэн үү?"}], "ideal": "халуун"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Gal haluun uu huiten uu?"}], "ideal": "Haluun"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Англид жолоо барихдаа аль талаа барьж явдаг вэ?"}], "ideal": "Зүүн"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Anglid joloo barihdaa ali talaa barij yavdag ve?"}], "ideal": "Zuun"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Зуны дараа ямар улирал болдог вэ?"}], "ideal": "Намар"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Zunii daraa yamar uliral boldog ve?"}], "ideal": "Namar"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Целсийн 0 грудасаас бага темпертурт байгаа усыг юу гэдэг вэ?"}], "ideal": "Мөс"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Tselsiin 0 gradusaas baga temperturt baigaa usiig yu gdg ve?"}], "ideal": ["Mus","Mos"]} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Хамгийн хөнгөн элемент юу вэ?"}], "ideal": "Гэли"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Hamgiin hungun element yu ve?"}], "ideal": ["Helium","Geli"]} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Наранд хамгийн ойрхон орших гараг юу вэ?"}], "ideal": "Буд"} {"input": [{"role": "system", "content": "You are OneWordGPT. Your answers should strictly only contain a single word and be in the same language and style as the user's question."}, {"role": "user", "content": "Narand hamgiin oirhon orshih garag yu ve?"}], "ideal": "Bud"} ``` </details>
- Loading branch information
