Update README.md

README.md

@@ -11,6 +11,7 @@ tags:
 - hybrid-corpus
 metrics:
 - accuracy
+- f1
 datasets:
 - anggars/mbti-emotion
 model-index:
@@ -24,8 +25,23 @@ model-index:
       type: anggars/mbti-emotion
     metrics:
     - type: accuracy
-      value: 0.8796
+      value: 0.9546
       name: Accuracy
+    - type: f1
+      value: 0.9544
+      name: F1 Macro
+language:
+  - en
+  - id
+  - ja
+  - ko
+  - es
+  - fr
+  - de
+  - zh
+  - pt
+  - ru
+  - it
 ---
 
 # XLM-RoBERTa Emotion (Domain-Adapted for Midwest Emo/Math Rock)
@@ -36,27 +52,25 @@ This model is a fine-tuned version of [xlm-roberta-base](https://huggingface.co/
 
 - **Model Type:** XLM-RoBERTa Base (Sequence Classification Head with 28 Nodes)
 - **Labels:** 28 Emotion Categories (e.g., sadness, grief, admiration, anger, joy)
-- **Dataset:** `anggars/mbti-emotion` (Hybrid Corpus: 112,351 synthetic narrative rows + 129 organic scraped lyrics)
+- **Dataset:** `anggars/mbti-emotion` (Hybrid Corpus: 120,060 total rows. Stratified split: 96,048 train / 24,012 eval)
 - **Language:** English & Indonesian (Multilingual)
-- **Training Environment:** Kaggle Compute (Dual Tesla T4 GPU, fp16 Mixed Precision)
+- **Training Environment:** Kaggle Compute (Dual NVIDIA Tesla T4 GPU, fp16 Mixed Precision)
 
 ## Architectural Innovations: Overcoming Domain Shift
 
-Initial iterations of this model were trained purely on synthetic narrative data, which caused severe **Domain Shift** when predicting real-world music lyrics (e.g., misclassifying depressive metaphors like *"drowning"* or *"heal this soul"* as *Admiration*). 
+Initial iterations of this model were trained purely on synthetic narrative data, which caused severe **Domain Shift** when predicting real-world music lyrics. To mitigate this blind spot, a **Hybrid Corpus Integration** was executed. The model was forced to adapt to organic lyrics scraped directly from Genius.com and augmented with high-quality, balanced synthetic data generated via Gemma-2B-IT. 
 
-To mitigate this blind spot, a **Hybrid Corpus Integration** was executed. The model was forced to adapt to organic lyrics scraped directly from Genius.com. The integration process successfully recalibrated the latent space, forcing the model to understand poetic contexts and lyrical structures. 
-
-*Note: The slight reduction in absolute accuracy (from previous baselines) and the increased validation loss are expected phenomena known as **Strategic Accuracy Drop** and **Softmax Calibration**. The aggressive weight decay (0.05) ensures the model does not overconfidently hallucinate on ambiguous lyrics, resulting in highly generalized, real-world zero-shot capabilities.*
+The integration process successfully recalibrated the latent space, forcing the model to understand poetic contexts and lyrical structures. The aggressive weight decay (0.05) ensures the model does not overconfidently hallucinate on ambiguous lyrics, resulting in highly generalized, real-world zero-shot capabilities.
 
 ## Training Results
 
-The following results were achieved on the evaluation set during the 3-epoch training process:
+The following results were achieved on the evaluation set (24,012 rows) during the 3-epoch training process:
 
-| Epoch | Step  | Validation Loss | Accuracy |
-|:-----:|:-----:|:---------------:|:--------:|
-| 1.0   | 5624  | 1.0018          | 0.8422   |
-| 2.0   | 11248 | 0.8367          | 0.8660   |
-| 3.0   | 16872 | 0.7716          | 0.8796   |
+| Epoch | Training Loss | Validation Loss | Accuracy | F1 Macro |
+|:-----:|:-------------:|:---------------:|:--------:|:--------:|
+| 1.0   | 0.2403        | 0.2086          | 0.9347   | 0.9339   |
+| 2.0   | 0.1403        | 0.1837          | 0.9480   | 0.9475   |
+| 3.0   | 0.1005        | 0.1888          | 0.9546   | 0.9544   |
 
 ## Intended Uses & Limitations
 
@@ -67,8 +81,6 @@ This model is explicitly designed for the backend NLP engine of music analytics
 
 ### Training Hyperparameters
 
-To ensure stable convergence on the complex organic lyrics and prevent overfitting on the synthetic data, the following hyperparameters were utilized:
-
 - **learning_rate:** 1.5e-05
 - **train_batch_size:** 16
 - **eval_batch_size:** 16
@@ -77,11 +89,4 @@ To ensure stable convergence on the complex organic lyrics and prevent overfitti
 - **optimizer:** AdamW with betas=(0.9,0.999) and epsilon=1e-08
 - **lr_scheduler_type:** linear
 - **num_epochs:** 3
-- **mixed_precision_training:** Native AMP (fp16)
-
-### Framework Versions
-
-- Transformers 4.44.2
-- Pytorch 2.5.1+cu124
-- Datasets 3.1.0
-- Tokenizers 0.20.3
+- **mixed_precision_training:** Native AMP (fp16)