Update README.md

README.md

@@ -149,8 +149,8 @@ The loss curve demonstrates stable convergence with the final training loss reac
 | **Benchmark** | **Metric** | **Base Gemma-3-27B-IT** | **LogicFlow-Gemma-3-27b-thinking** | **Improvement** |
 |---------------|------------|--------------------------|-------------------------------------|-----------------|
 | **Mathematical Reasoning** |
-| GSM8K | Exact Match | 82.6% | **89.5%** | **+6.9%** |
-| MATH | Accuracy | 50.0% | **76.8%** | **+26.8%** |
+| GSM8K | 5-shot | 82.6% | **89.5%** | **+6.9%** |
+| MATH | 5-shot | 50.0% | **76.8%** | **+26.8%** |
 | **Code Generation** |
 | MBPP | pass@1 | 65.6% | **69.0%** | **+3.4%** |
 | HumanEval | 0-shot | 48.8% | *Pending* | *TBD* |
@@ -158,9 +158,9 @@ The loss curve demonstrates stable convergence with the final training loss reac
 | IFEval | Prompt-level | *45.0%* | **40.0%** | **-5.0%** |
 | IFEval | Instruction-level | *58.0%* | **53.1%** | **-4.9%** |
 | **Advanced Mathematics** |
-| AIME25 | Problem Solving | ~8-12% | **13.3%** | **+1-5%** |
+| AIME25 | 5-shot | ~8-12% | **13.3%** | **+1-5%** |
 | **Scientific Reasoning** |
-| GPQA Diamond | Science QA | ~30-35% | **45.96%** | **+11-16%** |
+| GPQA Diamond | 5-shot | ~30-35% | **45.96%** | **+11-16%** |
 | **Knowledge & Understanding** |
 | MMLU | Overall Accuracy | 78.6% | **75.3%** | **-3.3%** |
 | MMLU STEM | Sciences & Math | ~70.0% | **71.6%** | **+1.6%** |