Update README.md

README.md

@@ -35,12 +35,6 @@ This model is the result of a 2-stage curriculum learning approach:
 - **Stage 1 (V1)**: Fine-tuned on 569k Norwegian NLI samples for semantic understanding
 - **Stage 2 (This model)**: Further fine-tuned on 103k Norwegian/Danish QA and paraphrase samples
 
-**Key Features:**
-- ✅ **Lowest overfitting**: Only 1.18% NDCG@10 degradation (vs 2.5-3.4% for other configurations)
-- ✅ **Decreasing eval loss**: -3.3% improvement during training (only configuration to improve generalization)
-- ✅ **Optimal hyperparameters**: Very low learning rate (5e-6), strong regularization, no warmup
-- ✅ **Production-ready**: Stable performance from step 500 onwards
-
 ## Training Details
 
 ### Stage 2 Training Configuration
@@ -126,17 +120,6 @@ This model is designed for:
 - **PAWS-X Norwegian**: 21,829 paraphrase pairs
 - **Supervised-DA**: 74,560 Danish sentence pairs
 
-## Evaluation & Comparison
-
-This model was selected from 4 Stage 2 variants based on overfitting analysis:
-
-| Variant | Best NDCG@10 | Degradation | Eval Loss Δ | Verdict |
-|---------|--------------|-------------|-------------|---------|
-| **This model (V3)** | **0.8781** | **1.18%** ✓ | **-3.3%** ✓ | **Winner** |
-| Original | 0.8750 | 2.52% | +8.2% ⚠️ | High eval loss increase |
-| Filtered | 0.9099 | 2.65% | +6.5% ⚠️ | Unfair dev set |
-| V2 (no warmup) | 0.8692 | 3.36% | +1.1% ⚠️ | Poor peak performance |
-
 ## Limitations
 
 - Optimized primarily for Norwegian text (with Danish/Swedish support)