Search papers, labs, and topics across Lattice.
The paper introduces Aryabhata 1.0, a 7B parameter language model specifically designed for math reasoning on the Indian JEE exam. The model is constructed via model merging and supervised fine-tuning with curriculum learning on high-quality CoT data generated through rejection sampling. Performance is further improved using reinforcement learning with verifiable rewards (RLVR) and novel exploration strategies like Adaptive Group Resizing and Temperature Scaling.
A 7B model can now outperform larger models on a challenging math exam, thanks to a novel combination of merging, curriculum learning, and reinforcement learning with verifiable rewards.
We present Aryabhata 1.0, a compact 7B parameter math reasoning model optimized for the Indian academic exam, the Joint Entrance Examination (JEE). Despite rapid progress in large language models (LLMs), current models often remain unsuitable for educational use. Aryabhata 1.0 is built by merging strong open-weight reasoning models, followed by supervised fine-tuning (SFT) with curriculum learning on verified chain-of-thought (CoT) traces curated through best-of-$n$ rejection sampling. To further boost performance, we apply reinforcement learning with verifiable rewards (RLVR) using A2C objective with group-relative advantage estimation along with novel exploration strategies such as Adaptive Group Resizing and Temperature Scaling. Evaluated on both in-distribution (JEE Main 2025) and out-of-distribution (MATH, GSM8K) benchmarks, Aryabhata outperforms existing models in accuracy and efficiency, while offering pedagogically useful step-by-step reasoning. We release Aryabhata as a foundation model to advance exam-centric, open-source small language models. This marks our first open release for community feedback (https://huggingface.co/PhysicsWallahAI/Aryabhata-1.0); PW is actively training future models to further improve learning outcomes for students.
