webAI Releases TwIL-LM: A 1.7B and 3B Formal-Logic Model Family for Autoformalization on Local Hardware
webAI released TwIL-LM, a family of 1.7B and 3B parameter models designed for autoformalization. These models translate English into first-order logic and run on local hardware with minimal VRAM requirements.
webAI has introduced TwIL-LM, a specialized model family aimed at bridging natural language and formal logic. The release includes two variants sized at 1.7 billion and 3 billion parameters, targeting the niche field of autoformalization.
Accessibility is a key feature of this release, as the models are designed to operate on local hardware. The larger 3B version requires only 4GB of VRAM or a CPU, while the smaller 1.7B model has a download size of approximately 1.06GB.
Functionally, these models translate English text into first-order logic and assess whether specific conclusions logically follow from given premises. This capability supports applications requiring rigorous logical verification rather than generative text completion.
The models are distributed under a non-commercial license, limiting their use to research or personal projects. This positioning suggests a focus on advancing open research in logical reasoning within language models.
This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.