Nemotron Research Reasoning Qwen 1.5B v2 RLVE

This model, developed by hamishivi, is a 1.5 billion parameter language model built upon the NVIDIA Nemotron-Research-Reasoning-Qwen-1.5B base. Its key differentiator is the application of RLVE (Reinforcement Learning with Verifiable Environments), a method designed to significantly improve reasoning and problem-solving capabilities. The model has a context length of 32768 tokens.

Key Capabilities

• Enhanced Reasoning: Demonstrates improved performance on challenging reasoning benchmarks such as AIME 2024/2025, OMEGA-500, and OlympiadBench.
• Problem Solving: Shows better results on BBEH and LiveCodeBench-v6, indicating stronger problem-solving skills.
• RLVE Optimization: Leverages a novel reinforcement learning approach to achieve superior analytical and logical deduction abilities compared to its base model.

Good for

• Complex Reasoning Tasks: Ideal for applications requiring advanced logical inference, mathematical problem-solving, and analytical thinking.
• Research and Development: Suitable for researchers exploring reinforcement learning techniques in language models and verifiable environments.
• Benchmarking: Can be used as a strong baseline or comparison model for evaluating new reasoning-focused LLM developments.

For more in-depth information on the RLVE method and training details, refer to the RLVE Paper and the RLVE GitHub Repository.