ma7583/pvs_oracle

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:7BQuant:FP8Context Size:4kPublished:Nov 24, 2025License:mitArchitecture:Transformer Open Weights Featherless Exclusive Cold

The ma7583/pvs_oracle is a 7 billion parameter language model developed by ma7583, specifically designed to assist in theorem proving within the PVS (Prototype Verification System) environment. This model acts as a command recommender, suggesting next steps in a proof based on the current sequent and previous commands. It is optimized for formal verification tasks, providing intelligent guidance for complex mathematical and logical proofs.

Loading preview...

PVS Oracle: AI-Assisted Theorem Proving

The ma7583/pvs_oracle is a 7 billion parameter language model developed by ma7583, specifically engineered to enhance theorem proving in the PVS (Prototype Verification System) environment. This model functions as a command recommender, suggesting appropriate next steps during a proof session. It is part of the broader PVSGym project, which aims to integrate LLMs into formal verification workflows.

Key Capabilities

  • Proof Step Recommendation: Given a current sequent and a history of up to three previous commands, the model predicts the most relevant next command to advance the proof.
  • Specialized Input Format: It processes highly structured input, including the current PVS sequent and prior proof commands, demonstrating sensitivity to newline formatting for accurate recommendations.
  • Integration with PVS: Designed to work in conjunction with PVS, providing intelligent assistance to users navigating complex formal proofs.

Use Cases

  • Accelerating Formal Verification: Helps PVS users by suggesting proof tactics, potentially speeding up the verification process.
  • Educational Tool: Can serve as a learning aid for new PVS users, guiding them through proof construction.
  • Research in AI for Theorem Proving: Provides a foundation for further research into applying large language models to automated and interactive theorem proving.

Technical Details

The model expects a specific input format for queries, including Current Sequent:, Prev Command 1:, Prev Command 2:, Prev Command 3:, and Next Command:. It can be deployed as a FastAPI web server for easy integration into other systems.