Multi-task Language Understanding on MMLU Biz
90F1 ScoreSharedRequest (Original)
Evaluation Results
| Method | Links | |
|---|---|---|
| SharedRequest (Original)Backbone LLM=GPT-4o2026.06 | 90 | |
| Non-privateBackbone LLM=GPT-4o2026.06 | 89.9 | |
| Non-privateBackbone LLM=GPT-4o-mini2026.06 | 85.3 | |
| SharedRequest (Original)Backbone LLM=GPT-4o-mini2026.06 | 85.1 | |
| SharedRequest (Simplified)epsilon=1, Backbone=GPT-4o2026.06 | 84.8 | |
| SharedRequest (Simplified)Backbone LLM=GPT-4o2026.06 | 84.8 | |
| SharedRequest (Simplified)epsilon=1, Backbone=GPT-4o-mini2026.06 | 81.7 | |
| SharedRequest (Simplified)Backbone LLM=GPT-4o-mini2026.06 | 81.7 | |
| InferDPTDP Type=Relaxed DP, epsilon=5, Backbone=GPT-4o2026.06 | 76.3 | |
| CusText+DP Type=Relaxed DP, epsilon=3, Backbone=GPT-4o2026.06 | 75.8 | |
| InferDPTDP Type=Relaxed DP, epsilon=3, Backbone=GPT-4o2026.06 | 75.8 | |
| CusText+DP Type=Relaxed DP, epsilon=5, Backbone=GPT-4o2026.06 | 75.4 | |
| InferDPTDP Type=Relaxed DP, epsilon=5, Backbone=GPT-4o-mini2026.06 | 74.6 | |
| CusText+DP Type=Relaxed DP, epsilon=5, Backbone=GPT-4o-mini2026.06 | 74.3 | |
| InferDPTDP Type=Relaxed DP, epsilon=1, Backbone=GPT-4o2026.06 | 71.2 | |
| InferDPTDP Type=Relaxed DP, epsilon=3, Backbone=GPT-4o-mini2026.06 | 70.8 | |
| InferDPTDP Type=Relaxed DP, epsilon=1, Backbone=GPT-4o-mini2026.06 | 70 | |
| CusText+DP Type=Relaxed DP, epsilon=3, Backbone=GPT-4o-mini2026.06 | 69.5 | |
| CusText+DP Type=Relaxed DP, epsilon=1, Backbone=GPT-4o2026.06 | 69.4 | |
| CusText+DP Type=Relaxed DP, epsilon=1, Backbone=GPT-4o-mini2026.06 | 68.6 | |
| Non-privateBackbone LLM=GPT-3.5-Turbo2026.06 | 67.1 | |
| SharedRequest (Original)Backbone LLM=GPT-3.5-Turbo2026.06 | 66.5 | |
| CusTextDP Type=Standard DP, epsilon=5, Backbone=GPT-4o2026.06 | 64.9 | |
| SharedRequest (Simplified)Backbone LLM=GPT-3.5-Turbo2026.06 | 63.8 | |
| CusTextDP Type=Standard DP, epsilon=5, Backbone=GPT-4o-mini2026.06 | 63.1 | |
| CusTextDP Type=Standard DP, epsilon=3, Backbone=GPT-4o2026.06 | 60.3 | |
| CusTextDP Type=Standard DP, epsilon=3, Backbone=GPT-4o-mini2026.06 | 55.3 | |
| DP-PromptDP Type=Standard DP, epsilon=5, Backbone=GPT-4o2026.06 | 54.2 | |
| DP-PromptDP Type=Standard DP, epsilon=3, Backbone=GPT-4o2026.06 | 53.3 | |
| DP-PromptDP Type=Standard DP, epsilon=5, Backbone=GPT-4o-mini2026.06 | 52.9 | |
| DP-PromptDP Type=Standard DP, epsilon=3, Backbone=GPT-4o-mini2026.06 | 52.2 | |
| CusTextDP Type=Standard DP, epsilon=1, Backbone=GPT-4o-mini2026.06 | 51.1 | |
| DP-PromptDP Type=Standard DP, epsilon=1, Backbone=GPT-4o-mini2026.06 | 49.7 | |
| DP-PromptDP Type=Standard DP, epsilon=1, Backbone=GPT-4o2026.06 | 49.6 | |
| CusTextDP Type=Standard DP, epsilon=1, Backbone=GPT-4o2026.06 | 47.3 | |
| RanTextDP Type=Standard DP, epsilon=5, Backbone=GPT-4o2026.06 | 43.5 | |
| RanTextDP Type=Standard DP, epsilon=5, Backbone=GPT-4o-mini2026.06 | 41.1 | |
| RanTextDP Type=Standard DP, epsilon=3, Backbone=GPT-4o2026.06 | 40.5 | |
| RanTextDP Type=Standard DP, epsilon=1, Backbone=GPT-4o2026.06 | 39 | |
| RanTextDP Type=Standard DP, epsilon=3, Backbone=GPT-4o-mini2026.06 | 38.8 | |
| RanTextDP Type=Standard DP, epsilon=1, Backbone=GPT-4o-mini2026.06 | 38.1 |