Humorous Image Captioning on Human Evaluation Dataset (test)
1,554.6Elo RatingHUMORCHAIN
Evaluation results
| Method | Links | ||
|---|---|---|---|
| HUMORCHAINStrategy=Theory-Guided Multi-Stage Reasoning2025.11 | 1,554.6 | 3.57 | |
| Method FStrategy=Few-shot + Rule-Guided + CoT2025.11 | 1,528.84 | 1 | |
| Method EStrategy=Rule-Guided + CoT2025.11 | 1,526.46 | 0.88 | |
| Method DStrategy=Few-shot + Rule-Based2025.11 | 1,505.43 | 0.73 | |
| Method CStrategy=Rule-Based2025.11 | 1,504.45 | 0.54 | |
| Method AStrategy=Zero-shot2025.11 | 1,498.09 | 0.52 | |
| Method BStrategy=Few-shot2025.11 | 1,495.05 | 0.46 | |
| OxfordTVG-HICStrategy=External Oxford University dataset based on humor-labeled caption pairs2025.11 | 1,467.4 | 0.55 | |
| CLoTStrategy=Sun Yat-sen University’s CLoT model2025.11 | 1,464.06 | 0.74 |