List your company

Metaphor Classification on MMSD Metaphor style 2.0 (test)

73.19AccuracyGemini 2.5 Flash
52.899658.167363.43568.7027Jan 23, 2026

Evaluation results

MethodLinks
Gemini 2.5 FlashPrompting=zero-shot2026.0173.190.76
SFT on Combined → GRPO (Combined)Prompting=zero-shot, Setup=Combined2026.0169.140.8
SFT on Combined → GRPO (Style specific)Prompting=zero-shot, Setup=Style specific2026.0168.190.79
LLaMA 90B Vision InstructPrompting=zero-shot2026.0167.560.76
Qwen2.5-VL 32B InstructPrompting=zero-shot2026.0161.310.69
Gemma 3 27BPrompting=zero-shot2026.0158.440.64
Phi-4 MultimodalPrompting=zero-shot2026.0155.710.55
LLaVA 1.5 7BPrompting=zero-shot2026.0153.680.56