Update model card with better formatting and related models

README.md

@@ -9,6 +9,8 @@ tags:
 - rag
 - fact-checking
 - token-classification
+- long-context
+- 32k
 datasets:
 - RAGTruth
 base_model: llm-semantic-router/modernbert-base-32k
@@ -23,7 +25,7 @@ model-index:
       name: Hallucination Detection
     dataset:
       type: RAGTruth
-      name: RAGTruth
+      name: RAGTruth Test Set
     metrics:
     - type: f1
       value: 77.49
@@ -33,7 +35,7 @@ model-index:
       name: Token-Level F1
 ---
 
-# ModernBERT-base-32k Hallucination Detector
+# 🥬 ModernBERT-base-32k Hallucination Detector
 
 A hallucination detection model fine-tuned on RAGTruth dataset using extended 32K context ModernBERT.
 
@@ -49,12 +51,21 @@ This model detects hallucinations in LLM-generated text by classifying each toke
 
 ## Performance
 
+Evaluated on **RAGTruth test set** (2,700 samples):
+
 | Metric | This Model | LettuceDetect BASE | LettuceDetect LARGE |
 |--------|------------|-------------------|---------------------|
-| **Example-Level F1** | **77.49%** | 75.99% | 79.22% |
+| **Example-Level F1** | **77.49%** ✅ | 75.99% | 79.22% |
 | Token-Level F1 | 51.47% | 56.27% | - |
+| Context Window | **32K** | 8K | 8K |
+
+### Key Results
+- ✅ **Beats LettuceDetect BASE** by +1.5% on example-level F1
+- ✅ **4x longer context** (32K vs 8K tokens)
+- ✅ **Same model size** as BASE (~150M parameters)
 
-**Beats LettuceDetect BASE** while supporting 4x longer context (32K vs 8K tokens).
+### Related Model
+- [`modernbert-base-32k-haldetect-combined`](https://huggingface.co/llm-semantic-router/modernbert-base-32k-haldetect-combined) - Trained on RAGTruth + HaluEval (48K samples)
 
 ## Usage