List your company
ResearchFormerly WizWand

Internal Humorous Image Captioning Dataset

85Win RateHUMORCHAIN (without discriminator) (J)
12.231.15068.9Nov 21, 2025
Updated 7mo ago

Evaluation results

MethodLinks
HUMORCHAIN (without discriminator) (J)Strategy=Theory-Guided Multi-Stage Reasoning (without discriminator), Total Samples=300, Evaluation Protocol=Pairwise comparison vs Zero-shot (A)2025.1185
HUMORCHAIN (without discriminator) (J)Strategy=Theory-Guided Multi-Stage Reasoning (without discriminator), Total Samples=300, Evaluation Protocol=Pairwise comparison vs Few-shot + Rule-Guided + CoT (F)2025.1183
HUMORCHAIN (I)Strategy=Theory-Guided Multi-Stage Reasoning, Total Samples=300, Evaluation Protocol=Pairwise comparison vs HUMORCHAIN (J)2025.1174.5
HUMORCHAIN (I)Strategy=Theory-Guided Multi-Stage Reasoning, Total Samples=300, Evaluation Protocol=Pairwise comparison vs Zero-shot (A)2025.1169.5
HUMORCHAIN (I)Strategy=Theory-Guided Multi-Stage Reasoning, Total Samples=300, Evaluation Protocol=Pairwise comparison vs Few-shot + Rule-Guided + CoT (F)2025.1168
Rule-Guided + CoT (E)Strategy=Adds Chain-of-Thought reasoning to theory-guided captioning, Total Samples=300, Evaluation Protocol=Pairwise comparison vs Rule-Based (C)2025.1153.5
Few-shot + Rule-Based (D)Strategy=Combines examples with theoretical references, Total Samples=300, Evaluation Protocol=Pairwise comparison vs Few-shot + Rule-Guided + CoT (F)2025.1152.3
Few-shot + Rule-Guided + CoT (F)Strategy=Combines all strategies without explicit multi-stage orchestration, Total Samples=300, Evaluation Protocol=Pairwise comparison vs Rule-Guided + CoT (E)2025.1151.8
Few-shot + Rule-Based (D)Strategy=Combines examples with theoretical references, Total Samples=300, Evaluation Protocol=Pairwise comparison vs Few-shot (B)2025.1150.8
Few-shot (B)Strategy=Example-based prompting with humor-style mimicry, Total Samples=300, Evaluation Protocol=Pairwise comparison vs Zero-shot (A)2025.1150.5
Zero-shot (A)Strategy=Direct image captioning without examples or theoretical cues, Total Samples=300, Evaluation Protocol=Pairwise comparison vs Rule-Based (C)2025.1150.2
Rule-Based (C)Strategy=References the four theories of humor, Total Samples=300, Evaluation Protocol=Pairwise comparison vs Zero-shot (A)2025.1149.8
Zero-shot (A)Strategy=Direct image captioning without examples or theoretical cues, Total Samples=300, Evaluation Protocol=Pairwise comparison vs Few-shot (B)2025.1149.5
Few-shot (B)Strategy=Example-based prompting with humor-style mimicry, Total Samples=300, Evaluation Protocol=Pairwise comparison vs Few-shot + Rule-Based (D)2025.1149.2
Rule-Guided + CoT (E)Strategy=Adds Chain-of-Thought reasoning to theory-guided captioning, Total Samples=300, Evaluation Protocol=Pairwise comparison vs Few-shot + Rule-Guided + CoT (F)2025.1148.2
Few-shot + Rule-Guided + CoT (F)Strategy=Combines all strategies without explicit multi-stage orchestration, Total Samples=300, Evaluation Protocol=Pairwise comparison vs Few-shot + Rule-Based (D)2025.1147.7
Rule-Based (C)Strategy=References the four theories of humor, Total Samples=300, Evaluation Protocol=Pairwise comparison vs Rule-Guided + CoT (E)2025.1146.5
Few-shot + Rule-Guided + CoT (F)Strategy=Combines all strategies, Total Samples=300, Evaluation Protocol=Pairwise comparison vs HUMORCHAIN (I)2025.1132
Zero-shot (A)Strategy=Direct image captioning, Total Samples=300, Evaluation Protocol=Pairwise comparison vs HUMORCHAIN (I)2025.1130.5
HUMORCHAIN (without discriminator) (J)Strategy=Theory-Guided Multi-Stage Reasoning (without discriminator), Total Samples=300, Evaluation Protocol=Pairwise comparison vs HUMORCHAIN (I)2025.1125.5
Few-shot + Rule-Guided + CoT (F)Strategy=Combines all strategies, Total Samples=300, Evaluation Protocol=Pairwise comparison vs HUMORCHAIN (J)2025.1117
Zero-shot (A)Strategy=Direct image captioning, Total Samples=300, Evaluation Protocol=Pairwise comparison vs HUMORCHAIN (J)2025.1115