ReTool SFT: Reinforcement Learning for Strategic Tool Use in LLMs

ReTool is a dataset for training large language models in strategic tool use via reinforcement learning. The dataset was created by JoeYing and was last updated on April 29, 2025. It is associated with experiments on the AIME2024 and AIME2025 benchmarks.

Use Cases

• Training LLMs for strategic tool selection based on reinforcement learning signals.
• Benchmarking tool-augmented reasoning performance on tasks like AIME2024 and AIME2025.
• Studying the convergence behavior of RL-based tool-use strategies compared to text-based approaches.

Strengths

• Dataset is associated with a specific research framework (ReTool) for strategic tool use.
• Last update was recent, on 2025-04-29 02:52:03.

Limitations

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

Provenance

Source

JoeYing on Hugging Face

Collection Method

Likely generated as part of the ReTool reinforcement learning research framework.

Freshness

Last updated 2025-04-29 02:52:03.