Tmax 15K: Reinforcement Learning Environment Instances for Language Model Training

AllenAI released a collection of roughly 15,000 reinforcement learning environment instances used to train the Tmax 9B and other Tmax language models. The dataset was published on Hugging Face and last updated on June 23, 2026. Specific details on generation and composition are documented in the associated research paper.

Use Cases

• Training language models via reinforcement learning based on the described RL environment instances.
• Fine-tuning instruction-following models based on the dataset's composition of prompts and responses.
• Benchmarking RLHF methodologies using the released collection of training instances.
• Studying the data generation and curation process for large-scale language model training.

Strengths

• Contains approximately 15,000 RL environment instances, providing a substantial training corpus.
• Served as the training data for the Tmax 9B model, indicating practical utility.
• Associated with a detailed research paper explaining its generation and makeup.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

AllenAI

Collection Method

Generated for training Tmax language models; specific methodology detailed in the associated paper.

Freshness

Last updated 2026-06-23 03:28:44; freshness should be verified.