List your company

Causal Reasoning on e-CARE n = 200

82.3P0 ScoreVernier
76.78878.21979.6581.081Jun 14, 2026

Evaluation results

MethodLinks
VernierBackbone=Qwen2.5-14B2026.0682.36121.3-1
Base (CLadder adapter)Backbone=Qwen2.5-14B2026.068260.521.5—
Base (CLadder adapter)Backbone=Qwen2.5-32B, Quantization=QLoRA2026.068262.519.5—
Base (CLadder adapter)Backbone=Qwen2.5-7B2026.0681.558.523—
VernierBackbone=Qwen2.5-32B, Quantization=QLoRA2026.0681.559.52213
VernierBackbone=Qwen2.5-7B2026.068162.218.8-18
Base (CLadder adapter)Backbone=Phi-3-mini (3.8B)2026.0679.55920.5—
VernierBackbone=Llama-3.1-8B2026.0679.361.517.8-26
Base (CLadder adapter)Backbone=Llama-3.1-8B2026.06795524—
Base (CLadder adapter)Backbone=Mistral-7B, Recipe=Default, lr=1e-42026.067955.523.5—
Base (CLadder adapter)Backbone=Mistral-7B, Recipe=Rescued, lr=5e-6, step=1502026.067955.523.5—
VernierBackbone=Phi-3-mini (3.8B)2026.0678.56018.5-10
VernierBackbone=Mistral-7B, Recipe=Default, lr=1e-42026.0677.55819.5-17
VernierBackbone=Mistral-7B, Recipe=Rescued, lr=5e-6, step=1502026.067756.520.5-13