List your company

Humor Classification on MMSD Humor style 2.0 (test)

80.71AccuracySFT on Combined → GRPO (Combined)
60.773265.949171.12576.3009Jan 23, 2026

Evaluation results

MethodLinks
SFT on Combined → GRPO (Combined)Prompting=zero-shot, Setup=Combined2026.0180.7193
Gemini 2.5 FlashPrompting=zero-shot2026.0179.1389
LLaMA 90B Vision InstructPrompting=zero-shot2026.0178.7187
SFT on Combined → GRPO (Style specific)Prompting=zero-shot, Setup=Style specific2026.0178.5991
Qwen2.5-VL 32B InstructPrompting=zero-shot2026.0177.1487
Phi-4 MultimodalPrompting=zero-shot2026.0170.3373
Gemma 3 27BPrompting=zero-shot2026.0162.2368
LLaVA 1.5 7BPrompting=zero-shot2026.0161.5461