Add placeholder link to code repository

README.md

@@ -1,20 +1,19 @@
 ---
+language:
+- en
 library_name: transformers
 license: other
 license_name: nvidia-open-model-license
-license_link: >-
-  https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/
+license_link: https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/
 pipeline_tag: text-generation
-language:
-  - en
 tags:
-  - nvidia
-  - reasoning
-  - math
-  - code
-  - supervised fine-tuning
-  - reinforcement learning
-  - pytorch
+- nvidia
+- reasoning
+- math
+- code
+- supervised fine-tuning
+- reinforcement learning
+- pytorch
 ---
 
 # AceReason-Nemotron 1.1: Advancing Math and Code Reasoning through SFT and RL Synergy
@@ -32,6 +31,8 @@ tags:
 
 We're thrilled to introduce [AceReason-Nemotron-1.1-7B](https://huggingface.co/nvidia/AceReason-Nemotron-1.1-7B), a math and code reasoning model built upon the Qwen2.5-Math-7B base. The model is first trained with supervised fine-tuning (SFT) on math and code tasks, then further enhanced through reinforcement learning (RL) using the same recipe as [AceReason-Nemotron-1.0-7B](https://huggingface.co/nvidia/AceReason-Nemotron-7B). We initiate RL training from various SFT models and find that stronger SFT models continue to produce consistently better results after large-scale RL, although the performance gap narrows during RL training. Thanks to its stronger SFT backbone, AceReason-Nemotron-1.1-7B significantly outperforms its predecessor and sets a record-high performance among Qwen2.5-7B-based reasoning models on challenging math and code reasoning benchmarks. For more details, check our [technical report](https://arxiv.org/abs/2506.13284).
 
+Code: TBD.
+
 ## Results
 
 We evaluate our model against competitive reasoning models of comparable size on AIME 2024, AIME 2025, and LiveCodeBench (LCB) v5 (2024/08/01 - 2025/02/01) and v6 (2025/02/01-2025/05/01). 
@@ -91,22 +92,53 @@ math_question = "MATH_QUESTION"
 math_instruction = "Please place your final answer inside \\boxed{}."
 system_instruction = "You are a helpful and harmless assistant. You should think step-by-step."
 
-final_prompt = "<|im_start|>system\n" + system_instruction + "<|im_end|>\n<|im_start|>user\n" + math_question + "\n\n" + math_instruction + "<|im_end|>\n<|im_start|>assistant\n<think>\n"
+final_prompt = "<|im_start|>system
+" + system_instruction + "<|im_end|>
+<|im_start|>user
+" + math_question + "
+
+" + math_instruction + "<|im_end|>
+<|im_start|>assistant
+<think>
+"
 ```
 3. We recommend using the following instruction for code questions:
 ```python
 code_question = "CODE_QUESTION"
 starter_code = "STARTER_CODE" # starter code function header, set empty string ("") if there is no starter code
 
-code_instruction_nostartercode = """Write Python code to solve the problem. Please place the solution code in the following format:\n```python\n# Your solution code here\n```"""
-code_instruction_hasstartercode = """Please place the solution code in the following format:\n```python\n# Your solution code here\n```"""
+code_instruction_nostartercode = """Write Python code to solve the problem. Please place the solution code in the following format:
+```python
+# Your solution code here
+```"""
+code_instruction_hasstartercode = """Please place the solution code in the following format:
+```python
+# Your solution code here
+```"""
 if starter_code != "":
-    code_question += "\n\n" + "Solve the problem starting with the provided function header.\n\nFunction header:\n" + "```\n" + starter_code + "\n```"
-    code_question += "\n\n" + code_instruction_hasstartercode
+    code_question += "
+
+" + "Solve the problem starting with the provided function header.
+
+Function header:
+" + "```
+" + starter_code + "
+```"
+    code_question += "
+
+" + code_instruction_hasstartercode
 else:
-    code_question += "\n\n" + code_instruction_nostartercode
+    code_question += "
 
-final_prompt = "<|im_start|>system\n" + system_instruction + "<|im_end|>\n<|im_start|>user\n" + code_question + "<|im_end|>\n<|im_start|>assistant\n<think>\n"
+" + code_instruction_nostartercode
+
+final_prompt = "<|im_start|>system
+" + system_instruction + "<|im_end|>
+<|im_start|>user
+" + code_question + "<|im_end|>
+<|im_start|>assistant
+<think>
+"
 ```
 4. Our inference engine for evaluation is vLLM==0.7.3 using top-p=0.95, temperature=0.6, max_tokens=32768.
 
@@ -134,4 +166,4 @@ June 16, 2025
   journal={arXiv preprint arXiv:2506.13284},
   year={2025}
 }
-```
+```