List your company

Semantic Consistency Evaluation on Semantic-stress diagnostic dataset (evaluation subset)

0.0093Mean ∆lexGemma3-4B
0.00450.03690.06930.1017Jun 7, 2026

Evaluation results

MethodLinks
Gemma3-4BL=27, N=9,0682026.060.00930.0092100-0.0016-0.001612.8
Qwen3VL-2BL=24, N=9,6612026.060.0190.0188100-0.0156-0.01585.3
InternVL3-2BL=23, N=9,7852026.060.03810.0377100-0.0399-0.04033.2
Qwen3VL-4BL=24, N=9,2262026.060.06810.0678100-0.0147-0.01498.6
Qwen3VL-8BL=32, N=9,6692026.060.12930.128499.90.02950.028885.3