| benchmark,reference_score,selected_score,delta | |
| math_reasoning,0.546,0.550,0.004 | |
| code_generation,0.646,0.650,0.004 | |
| text_classification,0.826,0.828,0.002 | |
| sentiment_analysis,0.792,0.792,0.000 | |
| question_answering,0.606,0.607,0.001 | |
| logical_reasoning,0.818,0.819,0.001 | |
| common_sense,0.736,0.736,0.000 | |
| reading_comprehension,0.698,0.700,0.002 | |
| dialogue_generation,0.643,0.644,0.001 | |
| summarization,0.766,0.767,0.001 | |
| translation,0.803,0.804,0.001 | |
| knowledge_retrieval,0.675,0.676,0.001 | |
| creative_writing,0.609,0.610,0.001 | |
| instruction_following,0.757,0.758,0.001 | |
| safety_evaluation,0.738,0.739,0.001 | |