Update README.md

README.md

@@ -21,7 +21,7 @@ tags:
 
 # 🦅 Supra-50M
 
-**Supra-50M** is a compact 50M-parameter causal language model built by SupraLabs, trained from scratch using a Llama-style architecture on 20 billion tokens of high-quality educational web text. Despite being significantly smaller than comparable open models, it achieves competitive or superior results on several key benchmarks. It is our first Supra Scalling Up plan model.
+**Supra-50M** is a compact 50M-parameter causal language model built by SupraLabs, trained from scratch using a Llama-style architecture on 20 billion tokens of high-quality educational web text. Despite being significantly smaller than comparable open models, it achieves competitive or superior results on several key benchmarks.
 
 ---
 
@@ -34,13 +34,13 @@ Supra-50M outperforms much larger models — GPT-2 Small (124M), SmolLM-135M, an
 ### Benchmark Table
 
 | Benchmark | Supra-50M *(ours)* | GPT-2 (124M) | SmolLM-135M | OpenELM-270M |
-|---|---|---|---|---|
-| **Parameters** | **50M** | 124M *(2.5× larger)* | 135M *(2.7× larger)* | 270M *(5.4× larger)* |
-| BLiMP (linguistics) | **76.3%** | ~63.0% | ~75.2% | ~68.0% |
-| SciQ (science) | **77.2%** | ~52.0% | ~74.5% | ~61.0% |
-| ARC-Easy (knowledge) | 52.2% | ~42.0% | **~55.0%** | ~46.0% |
-| PIQA (logic) | 62.2% | ~61.0% | **~63.3%** | ~60.5% |
-| HellaSwag (context) | 31.8% | ~31.0% | **~34.0%** | ~28.0% |
+| :--- | :--- | :--- | :--- | :--- |
+| **Parameters** | **50M** | 124M *(2.5×)* | 135M *(2.7×)* | 270M *(5.4×)* |
+| **BLiMP** (linguistics) | **76.3%** | 63.0% | **69.8%** | *(k.A.)* |
+| **SciQ** (science) | 77.2% | 53.2% | 73.4% | **84.70%** |
+| **ARC-Easy** (knowledge) | 52.2% | 42.0% | 49.2% | **45.08%** |
+| **PIQA** (logic) | 62.2% | 63.0% | 67.3% | **69.75%** |
+| **HellaSwag** (context) | 31.8% | 29.5% | 42.0% | **46.71%** |
 
 ---