Scientific Reasoning on GPQA (test)
11.44Error Rate (%)PAC
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| PACScore type=Verbalized score, LLM Backbone=Llama-based, Alpha=0.05, Epsilon=0.22026.01 | 11.44 | 0 | 42.52 | |
| PACScore type=Router-based score, LLM Backbone=Llama-based, Alpha=0.05, Epsilon=0.22026.01 | 11.89 | 0 | 39.67 | |
| G-PACScore type=Router-based score, LLM Backbone=Llama-based, Alpha=0.05, Epsilon=0.22026.01 | 11.9 | 0 | 39.88 | |
| G-PACScore type=Verbalized score, LLM Backbone=Llama-based, Alpha=0.05, Epsilon=0.22026.01 | 11.91 | 0 | 36.87 | |
| PACScore type=Logits-based score, LLM Backbone=Llama-based, Alpha=0.05, Epsilon=0.22026.01 | 12.38 | 7.03 | 18.13 | |
| G-PACScore type=Logits-based score, LLM Backbone=Llama-based, Alpha=0.05, Epsilon=0.22026.01 | 12.4 | 0 | 23.47 |