LH-Tech-AI/CritiqueCore_v1

1---2language: en3license: mit4library_name: transformers5tags:6- sentiment-analysis7- classification8- from-scratch9datasets:10- imdb11metrics:12- accuracy13model-index:14- name: CritiqueCore-v115  results:16  - task:17      type: text-classification18      name: Sentiment Analysis19    dataset:20      name: imdb21      type: imdb22    metrics:23    - type: accuracy24      value: 0.925pipeline_tag: text-classification26---27 28# CritiqueCore v129 30CritiqueCore v1 is a compact Transformer model trained **from scratch** for sentiment analysis. Unlike models that use transfer learning, this model was initialized with random weights and learned the nuances of language (including sarcasm and basic cross-lingual sentiment) exclusively from the IMDb movie reviews dataset.31 32## Model Description33- **Architecture:** Custom Mini-Transformer (DistilBERT-based configuration)34- **Parameters:** ~9.06 Million35- **Layers:** 236- **Attention Heads:** 437- **Hidden Dimension:** 25638- **Training Data:** IMDb Movie Reviews (25,000 samples)39- **Training Duration:** ~10 minutes on NVIDIA T4 GPU40 41## Capabilities42- **Sentiment Detection:** Strong performance on positive/negative English text.43- **Sarcasm Awareness:** Recognizes negative intent even when positive words are used (e.g., "CGI vomit").44- **Robustness:** Handles minor typos and maintains high confidence on structured feedback.45 46## Limitations47- **Domain Specificity:** Optimized for reviews. May struggle with complex multi-turn dialogues.48- **Multilingual:** While it shows some intuition for German, it was not explicitly trained on non-English data.49 50## How to use (Inference Script)51First, you have to download `CritiqueCore_v1_Model.zip` and unpack it. Then, you can use `inference.py` from this repos' files list. Have fun :D52 53## Examples54### Example 1: Standard movie review55**Input:**56```plaintext57This movie was an absolute masterpiece! The acting was incredible and I loved every second.58```59**Output:** POSITIVE (99.03% confidence)60 61### Example 2: Sarcasm62**Input:**63```plaintext64Oh great, another superhero movie. Just what the world needed. I loved sitting through 3 hours of CGI vomit.65```66**Output:** NEGATIVE (93.81% confidence)67 68### Example 3: Negative question69**Input:**70```plaintext71Why did they even produce it?72```73**Output:** NEGATIVE (99.37% confidence)74 75## Training code76The full training code can be found in this repo as `train.ipynb`.