This is a filtered version of the ACCORD dataset that limits the input-output prompt size to 20,000 tokens maximum. This filtering ensures compatibility with language models that have context length constraints while maintaining the diversity of combinatorial optimization problems.
The ACCORD (Algorithmic Combinatorial Optimization via Reinforcement Reasoning Data) dataset contains diverse instances of five combinatorial optimization… See the full description on the dataset page: https://huggingface.co/datasets/mideavalwisard/ACCORD_20K_token.
