List your company

Visual Reasoning on Hull-Bench

67.3AccuracyUnsilencing Latent Reasoning
58.990461.147763.30565.4623May 4, 2026

Evaluation results

MethodLinks
Unsilencing Latent ReasoningBackbone=Qwen2.5VL-7B2026.0567.3
LVRBackbone=Qwen2.5VL-7B2026.0566.67
DMLRBackbone=Qwen2.5VL-7B2026.0565.83
MonetBackbone=Qwen2.5VL-7B2026.0565.67
ICoTBackbone=Qwen2.5VL-7B2026.0565.51
VanillaBackbone=Qwen2.5VL-7B2026.0565.4
LVRRFBackbone=Qwen2.5VL-7B2026.0565.19
CCoTBackbone=Qwen2.5VL-7B2026.0564.88
ICoTBackbone=Qwen2.5VL-3B2026.0564.67
DMLRBackbone=Qwen2.5VL-3B2026.0564.67
CoVTBackbone=Qwen2.5VL-7B2026.0564.46
Unsilencing Latent ReasoningBackbone=Qwen2.5VL-3B2026.0564.35
VanillaBackbone=Qwen2.5VL-3B2026.0564.14
CCoTBackbone=Qwen2.5VL-3B2026.0564.03
MCoTBackbone=Qwen2.5VL-3B2026.0563.8
MCoTBackbone=Qwen2.5VL-7B2026.0563.62
LVRRFBackbone=Qwen2.5VL-3B2026.0560.99
LVRBackbone=Qwen2.5VL-3B2026.0559.31