List your company

Long-Context Reasoning on GPQA Diamond (out-of-distribution)

48.5AccuracyVTC-R1
25.41231.40637.443.394Jan 29, 2026

Evaluation results

MethodLinks
VTC-R1Architecture=Qwen3-VL-8B2026.0148.59.779.57
VTC-R1Architecture=Glyph2026.014610.736.96
SFTArchitecture=Glyph2026.0138.413.918.35
SFTArchitecture=Qwen3-VL-8B2026.0137.414.7826.88
TokenSkipArchitecture=Glyph2026.0135.915.459.93
Base SFTArchitecture=Glyph2026.0126.319.7414.43