Spaces:
Sleeping
Sleeping
import streamlit as st | |
st.set_page_config(page_title="HAERAE Open Research Questions", layout="wide") | |
st.title("HAERAE Open Research Questions") | |
st.write(""" | |
HAERAE is a non-profit research lab focused on the interpretability and evaluation of Korean language models. | |
Our mission is to advance the field with insightful benchmarks and tools. Below is an overview of our projects. | |
We've been doing most of our projects internally, but for those that have been unsolvable, | |
we are planning to open them to get help from the open-source community. | |
""") | |
st.header("HAERAE-Math Challenge") | |
st.write(""" | |
Today we are introducing our first challenge: HAERAE-Math. We've created high-quality instructions on math | |
but don't have an idea on how to generate high-quality answers for them. We are looking for solutions that | |
use open-source models with openly available licenses. | |
We have created a total of 20,000 instructions already and are generating more. We've opened up a preview | |
of 50 of them in this link: [HAERAE-Math Samples](https://huggingface.co/datasets/HAERAE-HUB/HAERAE-Math-samples) | |
For those who generate answers for the 50 and share the methodology/results with us, we'll share the | |
remaining instructions and credit for the resulting dataset. | |
""") | |
st.subheader("Example Question") | |
example_question = """ | |
νκ΅μ 보μ μ λ¬Έκ°κ° κ³ λνλ λ°μ΄ν° λ³΄νΈ μμ€ν μ κ°λ°νκ³ μμ΅λλ€. μ΄ μμ€ν μ 3μ°¨μ κΈ°ννμ μ κΈ λ©μ»€λμ¦μ μ¬μ©νλλ°, μ κΈ μ₯μΉλ μλΏ λͺ¨μμΌλ‘ λμ΄ μκ³ , λ°λ©΄μ λ°μ§λ¦μ 6cm, λμ΄λ 8cmμ λλ€. μ΄ μλΏ λͺ¨μμ μ κΈ μ₯μΉμλ μν΅ λͺ¨μμ μ΄μ κ° λ± λ§κ² λ€μ΄κ°κ² μ€κ³λμ΄ μμ΅λλ€. | |
보μ μ λ¬Έκ°λ λ λμ μμ€μ 보μμ μν΄ μν΅ λͺ¨μμ μ΄μ μμ ꡬ λͺ¨μμ μ κΈ μ₯μΉλ₯Ό μΆκ°νλ €κ³ ν©λλ€. μ΄ κ΅¬λ μν΅ μμ λ± λ€μ΄κ°λλ‘ μ€κ³λμ΄ μμ΅λλ€. | |
λ€μμ μ§λ¬Έλ€μ ν΄κ²°νμκΈ° λ°λλλ€: | |
1. μλΏ μμ λ± λ€μ΄κ°κ² μ€κ³λ μν΅μ λ°μ§λ¦μ μΌλ§μΈκ°μ? | |
2. μν΅ μμ λ± λ€μ΄κ°κ² μ€κ³λ ꡬμ λΆνΌλ μΌλ§μΈκ°μ? | |
3. μλΏ, μν΅, κ΅¬κ° λͺ¨λ κ°μ μ€μ¬μΆμ 곡μ νκ³ μμΌλ©° μλΏμ κΌλκΈ°μ κ³Ό μν΅, ꡬμ μ€μ¬μ μ΄ λμΌνλ€κ³ κ°μ νλ©΄, μλΏμμ μν΅μ΄ μ°¨μ§νλ λΉμ¨μ ꡬνμμ€. | |
4. μ΄μ μλΏμ λμ΄λ₯Ό 2λ°°λ‘ λ리μ. μλΏμ λμ΄κ° 16cmκ° λμμ λ, μν΅κ³Ό ꡬμ ν¬κΈ°μ λΆνΌλ μ΄λ»κ² λ³νλμ? | |
5. μλΏμ λμ΄μ λ°λ©΄μ λ°μ§λ¦μ κ°κ° hμ rμ΄λΌκ³ ν λ, μν΅κ³Ό ꡬμ μ΅λ λΆνΌλ₯Ό rκ³Ό hλ‘ νννμμ€. | |
μλΏ, μν΅, ꡬμ λΆνΌ 곡μμ μ¬μ©νμ¬ λ¬Έμ λ₯Ό ν΄κ²°νμκΈ° λ°λλλ€: | |
μλΏμ λΆνΌ: V = 1/3ΟrΒ²h | |
μν΅μ λΆνΌ: V = ΟrΒ²h | |
ꡬμ λΆνΌ: V = 4/3ΟrΒ³ | |
""" | |
st.code(example_question, language="markdown") | |
st.header("How to Participate") | |
st.write(""" | |
1. Access the 50 sample questions from the provided Hugging Face dataset link. | |
2. Generate high-quality answers for these questions using open-source models. | |
3. Document your methodology and results. | |
4. Share your findings with us through [contact information or submission form]. | |
5. If your approach is promising, we'll provide access to the full dataset of 20,000 instructions. | |
6. Collaborate with us to refine and improve the answer generation process. | |
7. Receive credit as a contributor to the final HAERAE-Math dataset. | |
""") | |
st.header("Why Participate?") | |
st.write(""" | |
- Contribute to advancing Korean language model research | |
- Gain access to a large, high-quality dataset of math instructions | |
- Collaborate with HAERAE researchers | |
- Receive recognition in the field of NLP and math education | |
- Potential for co-authorship on related publications | |
""") | |
st.header("Contact Us") | |
st.write(""" | |
For more information or to submit your results, please contact us at: | |
[Your contact information or a link to a submission form] | |
""") | |
st.sidebar.title("About HAERAE") | |
st.sidebar.info(""" | |
HAERAE is a non-profit research lab dedicated to advancing the field of | |
Korean language model interpretability and evaluation. Our work focuses on | |
creating insightful benchmarks and tools to push the boundaries of NLP research. | |
""") |