RL Dataset V2: Reinforcement Learning Data for Formal Theorem Proving

A reinforcement learning dataset for formal theorem proving, containing two task subsets: whole proof generation and self-revision. The dataset is associated with the Goedel-Prover-V2 paper and was uploaded by Goedel-LM to Hugging Face on March 2, 2026.

Use Cases

• Training reinforcement learning agents for automated proof generation based on the 'lean4_completion' subset.
• Developing self-correction mechanisms for formal proofs based on the 'lean4_revision' subset.
• Benchmarking AI performance on formal mathematics tasks based on the described task subsets.
• Studying the application of RL in code and logic synthesis for theorem proving.

Strengths

• Dataset is explicitly linked to a published research paper (Goedel-Prover-V2).
• Contains two distinct, task-oriented subsets for proof generation and revision.
• Last updated on March 2, 2026, indicating recent activity.

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, file formats, and license information are unknown, which may limit suitability assessment.

Provenance

Source

Goedel-LM on Hugging Face.

Collection Method

Likely generated for reinforcement learning experiments in formal theorem proving, as described in the associated paper.

Time Range

null

Freshness

Last updated 2026-03-02 19:27:49; freshness should be verified.

Geography

null