Tiny 1.7B TwIL-LM2 Beats Larger Models at Formal Logic Reasoning

TwIL-LM2, a 72M-parameter LoRA adapter on SmolLM2-1.7B that translates English into first-order logic for solver verification, outperforms far larger models like Qwen3-8B and Gemma-4-26B on strict formal reasoning benchmarks.

2026-08-22 ~ 2026-08-23 · 2 related posts

• 1.7B Model Outperforms Qwen3-8B in Strict Formal Logic Reasoning — Creative-Fig522 · 2026-08-22
• TwIL-LM2 1.7B outperforms larger models in strict formal reasoning — HopefulMode429 · 2026-08-23