Multi-label food category extraction on HFPP database
95.39Grow AccuracyRole-based Framework (Llama-3.3)
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Role-based Framework (Llama-3.3)Prompting Strategy=Role-based, Backbone=Llama-3.32026.04 | 95.39 | 93.32 | 90.78 | 95.16 | 69.82 | 63.13 | |
| Few-ShotPrompting Strategy=Few-Shot2026.04 | 94.49 | 93.16 | 88.25 | 95.16 | 69.35 | 64.97 | |
| Chain-of-ThoughtPrompting Strategy=Chain-of-Thought2026.04 | 94.49 | 94.95 | 90.37 | 78.21 | 68.81 | 63.3 | |
| Zero-ShotPrompting Strategy=Zero-Shot2026.04 | 94.04 | 93.35 | 80.96 | 88.07 | 71.1 | 36.01 | |
| Role-based Framework (Qwen-3)Prompting Strategy=Role-based, Backbone=Qwen-32026.04 | 65.95 | 96.22 | 95.56 | 96.05 | 72.2 | 49.34 |