Improve model card with abstract, GitHub link, and comprehensive download section

This pull request aims to enhance the model card for `Fast-Math-R1-14B` by:
* Adding the paper abstract for a more immediate and detailed understanding of the model.
* Introducing a prominent "Code:" link to the GitHub repository for easier access to the project's codebase.
* Replacing the current "Dataset" section with a more comprehensive "Download" section, sourced from the original GitHub repository, which includes links to the model itself and related models, in addition to the datasets.

These updates improve the discoverability of associated resources and the overall completeness of the model card.

README.md

@@ -1,15 +1,6 @@
 ---
 base_model:
 - deepseek-ai/DeepSeek-R1-Distill-Qwen-14B
-license: apache-2.0
-pipeline_tag: text-generation
-library_name: transformers
-tags:
-- math
-- reasoning
-- llm
-- mathematical-reasoning
-- aimo
 datasets:
 - RabotniKuma/Fast-Math-R1-SFT
 - RabotniKuma/Fast-Math-R1-GRPO
@@ -18,14 +9,28 @@ datasets:
 - qihoo360/Light-R1-SFTData
 language:
 - en
+library_name: transformers
+license: apache-2.0
 metrics:
 - pass@1
+pipeline_tag: text-generation
+tags:
+- math
+- reasoning
+- llm
+- mathematical-reasoning
+- aimo
 ---
 
 # Kaggle AI Mathematical Olympiad - Progress Prize 2 - 9th Place Solution (Fast-Math-R1-14B)
 
 This model was presented in the paper [A Practical Two-Stage Recipe for Mathematical LLMs: Maximizing Accuracy with SFT and Efficiency with Reinforcement Learning](https://huggingface.co/papers/2507.08267).
 
+## Abstract
+Enhancing the mathematical reasoning of Large Language Models (LLMs) is a pivotal challenge in advancing AI capabilities. While Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) are the dominant training paradigms, a systematic methodology for combining them to maximize both accuracy and efficiency remains largely unexplored. This paper introduces a practical and effective training recipe that strategically integrates extended SFT with RL from online inference (GRPO). We posit that these methods play complementary, not competing, roles: a prolonged SFT phase first pushes the model's accuracy to its limits, after which a GRPO phase dramatically improves token efficiency while preserving this peak performance. Our experiments reveal that extending SFT for as many as 10 epochs is crucial for performance breakthroughs, and that the primary role of GRPO in this framework is to optimize solution length. The efficacy of our recipe is rigorously validated through top-tier performance on challenging benchmarks, including a high rank among over 2,200 teams in the strictly leak-free AI Mathematical Olympiad (AIMO). This work provides the community with a battle-tested blueprint for developing state-of-the-art mathematical reasoners that are both exceptionally accurate and practically efficient. To ensure full reproducibility and empower future research, we will open-source our entire framework, including all code, model checkpoints, and training configurations at this https URL .
+
+Code: [https://github.com/analokmaus/kaggle-aimo2-fast-math-r1](https://github.com/analokmaus/kaggle-aimo2-fast-math-r1)
+
 ## Team
 - Hiroshi Yoshihara @ [Aillis Inc.](https://aillis.jp/en), [The Univ. of Tokyo](https://publichealth.f.u-tokyo.ac.jp/#page_home)
 - Yuichi Inoue @ [Sakana AI](https://sakana.ai)
@@ -37,7 +42,7 @@ which achieves up to 60% (on average approx. 30%) faster inference while maintai
 
 In addition, we trained and open-sourced `Fast-OpenMath-Nemotron-14B`, an efficiency-optimized version of NVIDIA’s `OpenMath-Nemotron-14B`, following the same approach.
 
-Technical details can be found in [Kaggle Discussion](https://www.kaggle.com/competitions/ai-mathematical-olympiad-progress-prize-2/discussion/571252) and [Github](https://github.com/analokmaus/kaggle-aimo2-fast-math-r1).
+Technical details can be found in [Kaggle Discussion](https://www.kaggle.com/competitions/ai-mathematical-olympiad-progress-prize-2/discussion/571252).
 
 ## Evaluation
 <img src="https://github.com/analokmaus/kaggle-aimo2-fast-math-r1/blob/master/assets/pass1_aime_all.png?raw=true" max-height="400px">
@@ -87,9 +92,13 @@ Technical details can be found in [Kaggle Discussion](https://www.kaggle.com/com
 |                     | 12000        | 65.1             | 7775               | 49.4             | 8733               | 
 |                     | 8000         | 50.7             | 6260               | 36               | 6618               | 
 
-## Dataset
-- [Our first stage SFT dataset](https://huggingface.co/datasets/RabotniKuma/Fast-Math-R1-SFT)
-- [Our second stage GRPO dataset](https://huggingface.co/datasets/RabotniKuma/Fast-Math-R1-GRPO)
+## Download
+- `Fast-Math-R1-14B` model is available at [Huggingface](https://huggingface.co/RabotniKuma/Fast-Math-R1-14B) and [Kaggle Models](https://www.kaggle.com/models/analokamus/fast_math_r1_14b/).
+- `Fast-OpenMath-Nemotron-14B` model is available at [Huggingface](https://huggingface.co/RabotniKuma/Fast-OpenMath-Nemotron-14B)
+- `Fast-Math-Qwen3-14B` model is available at [Huggingface](https://huggingface.co/RabotniKuma/Fast-Math-Qwen3-14B)
+- [First stage SFT dataset](https://huggingface.co/datasets/RabotniKuma/Fast-Math-R1-SFT)
+- [Second stage GRPO dataset](https://huggingface.co/datasets/RabotniKuma/Fast-Math-R1-GRPO)
+- (Optional) [Token scheduler dataset](https://huggingface.co/datasets/RabotniKuma/Fast-Math-R1-Token-Scheduler)
 
 ## Inference
 ### vLLM
@@ -196,26 +205,26 @@ wait
 ## Technical details
 Detailed report is available on [Kaggle Disucussion](https://www.kaggle.com/competitions/ai-mathematical-olympiad-progress-prize-2/discussion/571252).
 
-### First stage: intensive SFT using a high-difficulty dataset
-#### Dataset
+## First stage: intensive SFT using a high-difficulty dataset
+### Dataset
 - [OpenR1 Math](https://huggingface.co/datasets/open-r1/OpenR1-Math-220k): We randomly sampled 3000 examples where the R1’s trace had more than 12800 tokens and an accuracy of over 50%, along with another 3000 examples where the accuracy ranged between 50% and 75%.
-- [openr1_hard](https://huggingface.co/datasets/hoanganhpham/openr1_hard):  "~2.5k hard samples from open-r1-math-220k. Samples deemed as hard were unsolvable by r1-distill-32b after 4 tries."
+- [openr1_hard](https://huggingface.co/datasets/hoanganhpham/openr1_hard): "~2.5k hard samples from open-r1-math-220k. Samples deemed as hard were unsolvable by r1-distill-32b after 4 tries."
 - [Light-R1-SFTData](https://huggingface.co/datasets/qihoo360/Light-R1-SFTData): We used the 2nd stage data from Light-R1-SFTData.
 
 We merged all the datasets mentioned above, removed duplicates, and selected the correct generation with the shortest token length. For samples in the Light-R1 dataset where ground truth answers were not provided, we extracted and substituted the answers from the R1 traces. As a result, we constructed a **high-difficulty dataset consisting of 7900 problem - R1 trace - answer sets**.
 
 [Our first stage SFT dataset](https://huggingface.co/datasets/RabotniKuma/Fast-Math-R1-SFT)
 
-#### Training
+### Training
 A full-parameter supervised fine-tuning training was conducted on a machine with 8 H200 GPUs, using the SFTTrainer from the trl library.
 
-### Second stage: GRPO for more efficient reasoning
-#### Dataset
+## Second stage: GRPO for more efficient reasoning
+### Dataset
 - [Light-R1-SFTData](https://huggingface.co/datasets/qihoo360/Light-R1-SFTData): We extracted the answers from the 2nd stage SFT data of Light-R1.
 
 [Our second stage GRPO dataset](https://huggingface.co/datasets/RabotniKuma/Fast-Math-R1-GRPO)
 
-#### Training
+### Training
 We used the [faster implementation of trl GRPOTrainer](https://github.com/nhannguyen2709/open-r1).
 
 Reward functions: