Update README.md

README.md

@@ -1,4 +1,30 @@
-# AffectDynamics(Team AGI) — Longitudinal Affect Prediction Model
+---
+language:
+- en
+
+license: apache-2.0
+
+tags:
+- semeval
+- semeval-2026
+- emotion
+- affect-prediction
+- temporal-nlp
+- transformers
+- roberta
+
+datasets:
+- semeval
+
+pipeline_tag: text-classification
+
+library_name: transformers
+
+metrics:
+- pearson-correlation
+---
+
+# AffectDynamics (Team AGI) — Longitudinal Affect Prediction Model
 
 AffectDynamics is a temporal affect modeling system developed for **SemEval-2026 Task 2: Predicting Variation in Emotional Valence and Arousal over Time from Ecological Essays**.
 
@@ -22,17 +48,18 @@ The system consists of four main components:
 - Produces contextual embeddings for each text input.
 
 Different pooling strategies are used depending on text type:
-- Essays → CLS / pooler representation
-- Feeling word lists → mean pooled token embeddings
+
+- Essays → CLS / pooler representation  
+- Feeling word lists → mean pooled token embeddings  
 
 ### 2. Temporal Encoder
 - **Unidirectional GRU**
-- Models longitudinal emotional dynamics across user timelines.
-- Ensures **causal temporal modeling** (no future information leakage).
+- Models longitudinal emotional dynamics across user timelines
+- Ensures **causal temporal modeling** (no future information leakage)
 
 ### 3. User Conditioning
 - **Gated user embedding**
-- Incorporates user-level statistics such as:
+- Uses user statistics such as:
   - number of samples
   - timeline length
   - emotional entropy
@@ -40,7 +67,6 @@ Different pooling strategies are used depending on text type:
 This allows interpolation between **user-specific** and **global representations**.
 
 ### 4. Prediction Heads
-The model supports three prediction tasks:
 
 | Task | Description |
 |-----|-------------|
@@ -56,10 +82,10 @@ The model was trained using the official **SemEval-2026 Task 2 dataset**.
 
 ### Dataset statistics
 
-- **Total texts:** 5,285
-- **Training texts:** 2,764
-- **Users:** 182 total (137 in training)
-- **Time span:** 2021–2024
+- Total texts: **5,285**
+- Training texts: **2,764**
+- Users: **182 total (137 training users)**
+- Time span: **2021–2024**
 
 Each entry contains:
 
@@ -72,7 +98,7 @@ Each entry contains:
 | valence | Emotional valence (-2 to 2) |
 | arousal | Emotional arousal (0 to 2) |
 
-The texts were written by **U.S. service-industry workers** describing how they felt at the moment.
+The texts were written by **U.S. service-industry workers** describing their emotional state.
 
 ---
 
@@ -82,10 +108,10 @@ The texts were written by **U.S. service-industry workers** describing how they
 
 - Optimizer: **AdamW**
 - Scheduler: **OneCycleLR**
-- Batch size: 4
-- Training epochs: 10
+- Batch size: **4**
+- Training epochs: **10**
 
-### Learning rates
+### Learning Rates
 
 | Component | Learning Rate |
 |----------|---------------|
@@ -98,14 +124,14 @@ The texts were written by **U.S. service-industry workers** describing how they
 | Task | Loss |
 |----|----|
 | Subtask 1 | Ordinal regression with label smoothing |
-| Subtask 2A | Smooth L1 loss for delta prediction |
+| Subtask 2A | Smooth L1 loss |
 | Subtask 2B | Mean squared error |
 
 ---
 
 # Evaluation Results
 
-Official evaluation results from SemEval-2026 Task 2:
+Official evaluation results from **SemEval-2026 Task 2**:
 
 | Task | Metric | Valence | Arousal |
 |----|----|----|----|
@@ -119,30 +145,28 @@ The model demonstrates strong performance on **absolute affect prediction**, but
 
 # Intended Use
 
-This model is intended for **research purposes** including:
+This model is intended for **research purposes**, including:
 
-- longitudinal affect modeling
-- emotion prediction from text
-- temporal NLP modeling
-- ecological momentary assessment analysis
+- longitudinal affect modeling  
+- emotion prediction from text  
+- temporal NLP modeling  
+- ecological momentary assessment analysis  
 
 ---
 
 # Limitations
 
-Several limitations should be considered:
-
 1. **Stability bias**
-   - Temporal modeling tends to smooth predictions, reducing sensitivity to abrupt emotional changes.
+   - Temporal modeling smooths predictions and reduces sensitivity to abrupt changes.
 
 2. **Dataset domain**
-   - Data comes from a specific population (U.S. service-industry workers), which may limit generalization.
+   - Data originates from a specific population (U.S. service-industry workers).
 
-3. **Small number of users**
-   - Training data includes only 137 users.
+3. **Limited users**
+   - Only **137 users** in training data.
 
 4. **Change prediction difficulty**
-   - Predicting emotional deltas is significantly harder than predicting absolute states.
+   - Predicting emotional deltas is harder than predicting absolute states.
 
 ---
 
@@ -152,23 +176,26 @@ Emotion prediction models must be used responsibly.
 
 Potential concerns include:
 
-- **Privacy risks** when modeling personal emotional data
-- **Misuse for emotional manipulation**
-- **Bias from dataset demographics**
+- privacy risks from modeling personal emotional data  
+- misuse for manipulation or surveillance  
+- dataset demographic bias  
 
-This model should **not be used for clinical or psychological diagnosis**.
+This model **should not be used for clinical or psychological diagnosis**.
 
 ---
 
 # Reproducibility
 
-Code and training pipeline are available at:
+Code and training pipeline:
 
-**GitHub Repository**  
 https://github.com/ezylopx5/AffectDynamics-SemEval2026Task2
 
+Model weights:
+
+https://huggingface.co/Haxxsh/AffectDynamics-SemEval2026Task2
+
 ---
 
 # Citation
 
-If you use this model, please cite the system description paper:
+If you use this model, please cite: