ThuraAung1601/Qwen2.5-Coder-0.5B-Instruct_reform-dafny-loop-inv-gen_GRPO-full

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:0.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Oct 2, 2026License:cc-by-nc-4.0Architecture:Transformer Open Weights Featherless Exclusive Cold

ThuraAung1601/Qwen2.5-Coder-0.5B-Instruct_reform-dafny-loop-inv-gen_GRPO-full is a 0.5 billion parameter model, fine-tuned from Qwen/Qwen2.5-Coder-0.5B-Instruct, specifically for Dafny formal verification tasks. It excels at automatically adding loop invariants to Dafny programs to ensure verification, achieving 51.02% pass@1 on DafnyBench-cleaned across 733 tasks. This model's primary use case is generating verifiable Dafny code by completing missing loop invariants without altering other program elements.

Loading preview...

Overview

This model, ThuraAung1601/Qwen2.5-Coder-0.5B-Instruct_reform-dafny-loop-inv-gen_GRPO-full, is a specialized 0.5 billion parameter language model fine-tuned from the Qwen2.5-Coder-0.5B-Instruct base. Its core function is to assist in Dafny formal verification by automatically generating and inserting loop invariants into Dafny programs, enabling them to pass verification.

Key Capabilities

  • Automated Loop Invariant Generation: The model is specifically trained to add correct loop invariants to Dafny programs.
  • Dafny Program Verification: Its training objective, utilizing GRPO with rewards from the Dafny verifier, ensures the generated invariants lead to verifiable code.
  • High Performance on DafnyBench: Achieves a 51.02% pass@1 rate on the DafnyBench-cleaned dataset, covering 733 tasks.
  • Program Completion: Outputs the complete Dafny program with invariants, without altering existing code or specifications.

Good for

  • Dafny Formal Verification: Ideal for developers and researchers working with Dafny who need assistance in proving program correctness.
  • Automating Invariant Discovery: Useful for reducing the manual effort involved in writing complex loop invariants.
  • Educational Purposes: Can serve as a tool for understanding how loop invariants are constructed to satisfy Dafny's verifier.

Usage

The model expects a system message defining its role as a Dafny expert and a user message containing the Dafny program within a dafny code block.

Finetunes
1 models