Update README.md

README.md

@@ -17,23 +17,16 @@ tags:
 ---
 # mCoT: Multilingual Instruction Tuning for Reasoning Consistency in Language Models
 
-Paper: []()
+Paper: https://arxiv.org/abs/2406.02301
 
-Code: []()
+Code: https://github.com/laihuiyuan/mCoT
 
-Dataset: []()
+Dataset: https://huggingface.co/datasets/laihuiyuan/mCoT-MATH
 
 ### Introduction
-We introduce mCoT-MATH, a 7B parameter model for multilingual math reasoning, which achieves impressive consistency across languages. mCoT is based on [Mistral-7B](https://huggingface.co/mistralai/Mistral-7B-v0.1) and trained on mCoT-MATH, the first large-scale multilingual math CoT reasoning dataset containing around 6.3 million samples for 11 diverse languages.
+We introduce mCoT, a 7B parameter model for multilingual math reasoning that achieves impressive multilingual reasoning consistency across multiple languages. 
+Based on [Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1), mCoT is trained on [mCoT-MATH](https://huggingface.co/datasets/laihuiyuan/mCoT-MATH), the first large-scale multilingual math CoT reasoning dataset containing around 6.3 million samples for 11 diverse languages.
 
-### 🤗 Dataset: [mCoT-MATH](https://huggingface.co/datasets/laihuiyuan/mCoT-MATH)
-
-Based on [MetaMathQA](https://github.com/meta-math/MetaMath) and [MathInstruct](https://github.com/TIGER-AI-Lab/MAmmoTH)
-, we compile [mCoT-MATH](https://huggingface.co/datasets/laihuiyuan/mCoT-MATH) using machine translation.
-
-| Language  | SW    | BN    | TE    | TH    | JA    | ZH    | RU    | ES    | FR    | DE    | DE    |Overall |
-|:----------|:------|:------|:------|:------|:------|:------|:------|:------|:------|:------|:------|--------|
-| mCoT-MATH | ~580K | ~580K | ~580K | ~580K | ~580K | ~580K | ~580K | ~580K | ~580K | ~580K | ~580K | ~6.3M  |
 
 ### Results on [MGSM](https://arxiv.org/abs/2210.03057v1)
 | Language               | SW   | BN   | TE   | TH   | JA   | ZH   | RU   | ES   | FR   | DE   | EN   | 
@@ -68,7 +61,10 @@ Based on [MetaMathQA](https://github.com/meta-math/MetaMath) and [MathInstruct](
 
 ### Prompt Template
 ```bash
-# Language
+# Template
+template = "Question: \n{question} \nAnswer: \n{language}\n"
+
+# Language prompt
 bn = "আসুন ধাপে ধাপে চিন্তা করি।"
 de = "Denken wir Schritt für Schritt."
 en = "Let's think step by step."
@@ -81,10 +77,11 @@ te = "అంచెలంచెలుగా ఆలోచిద్దాం."
 th = "ลองคิดทีละขั้นตอน"
 zh = "让我们一步步思考。"
 
-# Math Question
-math = "A robe takes 2 bolts of blue fiber and half that much white fiber.  How many bolts in total does it take?"
+# Math question
+math_en = "A robe takes 2 bolts of blue fiber and half that much white fiber.  How many bolts in total does it take?"
 
-Prompt = "Question: \n[Math Question] \nAnswer: \n[Language]\n[CoT Reasoning]"
+# An example for the English question
+prompt = template.format(question=math_en, language=en)
 ```
 
 ### Citation