Training Step-Level Reasoning Verifiers with Formal Verification Tools
Paper
•
2505.15960
•
Published
•
7
FoVer
Process Reward Models (PRMs) trained on step-level error labels automatically annotated by formal verification tools.
-
-
ryokamoi/Llama-3.1-8B-FoVer-PRM
Text Generation • 8B • Updated • 25 -
ryokamoi/Qwen-2.5-7B-FoVer-PRM
Text Generation • 8B • Updated • 12 • 1 -
ryokamoi/FoVer-FormalLogic-Llama-3.1-8B
Viewer • Updated • 10.7k • 65 -
ryokamoi/FoVer-FormalLogic-Qwen-2.5-7B
Viewer • Updated • 10.7k • 107 -
ryokamoi/FoVer-FormalProof-Llama-3.1-8B
Viewer • Updated • 10.7k • 78 -
ryokamoi/FoVer-FormalProof-Qwen-2.5-7B
Viewer • Updated • 10.7k • 100 -
ryokamoi/FoVer-FormalLogic-FormalProof-Llama-3.1-8B-LastStepBalanced-40k
Viewer • Updated • 40k • 92 -
ryokamoi/FoVer-FormalLogic-FormalProof-Qwen-2.5-7B-LastStepBalanced-40k
Viewer • Updated • 40k • 70 -
ryokamoi/FoVer-misc
Updated • 137