XGBoost + Engineered Features (Misinformation Detection in Engineering)
This model is an XGBoost classifier trained on a 12-dimensional engineered feature set for detecting AI-generated misinformation in engineering documents. It does not use Transformer embeddings, relying entirely on structured features.
Class 0: Real engineering documents
Class 1: AI-generated misinformation
Model Components
xgb_model.json → Serialized XGBoost model weights
scaler.pkl → Scikit-learn StandardScaler for feature normalization
Training Details
Features (12D):
Document length (characters, words, sentences)
Punctuation density
Readability proxies (Flesch–Kincaid, Gunning Fog)
Engineering/safety keyword ratios
Numeric feature counts (e.g., decimals, scientific notation)
Standard acronym detection (ISO, IEEE, ANSI, etc.)
Framework: XGBoost (tree-based gradient boosting)
Training split: EMC dataset (train/val/test)
Evaluation metric: Macro F1
Intended Uses
Lightweight detection of AI-gener…
Gated Fusion (XLM-RoBERTa + Engineered Features with Dynamic Gating)
This model is a fusion architecture that combines XLM-RoBERTa embeddings with engineered linguistic and domain-specific features using a learned gating mechanism. It was trained for misinformation detection in engineering documents.
Class 0: Real engineering documents
Class 1: AI-generated misinformation
Model Components
fusion_gated.pt → PyTorch model weights
scaler.pkl → Scikit-learn scaler for the 12-dimensional engineered feature set
Base encoder: xlm-roberta-base (mean-pooled hidden states)
Training Details
Fusion Mechanism: Feature vectors are projected into a latent space and combined with Transformer embeddings through a dynamic gate, which learns how much weight to assign to engineered features vs. text embeddings.
Engineered Features (12D): Counts, readability proxies, punctuation density, engineering/safety keyword ratios, and numeric/standards indicators.
Optimizer: AdamW
Sequence length: 256 t…
Simple Fusion (XLM-RoBERTa + Engineered Features)
This model is a fusion architecture that combines XLM-RoBERTa embeddings with engineered linguistic and domain-specific features for misinformation detection in engineering texts.
Class 0: Real engineering documents
Class 1: AI-generated misinformation
Model Components
fusion_simple.pt → PyTorch model weights
scaler.pkl → Scikit-learn scaler for the 12-dimensional engineered feature set
Base encoder: xlm-roberta-base (mean-pooled hidden states)
Training Details
Fusion Mechanism: Naive concatenation of Transformer embeddings with scaled engineered features.
Engineered Features (12D): Include counts, readability proxies, punctuation density, engineering/safety keyword ratios, and numeric/standards signals.
Optimizer: AdamW
Sequence length: 256 tokens
Datasets: EMC Dataset
Intended Uses
Detecting AI-generated misinformation in engineering documents.
Serving as a hybrid baseline for comparing Transformer-only vs. feature-onl…
XLM-RoBERTa (Text-Only) Fine-Tuned on EMC
This model is a fine-tuned version of xlm-roberta-base trained on the Engineering Misinformation Corpus (EMC).
It was optimized for binary classification:
Class 0: Real engineering documents
Class 1: AI-generated misinformation
Unlike the fusion models, this checkpoint uses only raw text inputs (no engineered features).
It serves as a strong baseline for end-to-end Transformer performance on technical misinformation detection.
Training Details
Base model: xlm-roberta-base
Objective: Binary classification (cross-entropy)
Sequence length: 256 tokens
Optimizer: AdamW
Datasets: EMC Dataset
(train/val/test splits provided)
Intended Uses
Detecting AI-generated misinformation in engineering and technical texts.
Benchmarking Transformer-only models against feature-based or hybrid models.
⚠️ Limitations:
Trained only on engineering-domain corpora.
Performance may degrade on general-domain misinformation tasks.
Does not incorporate nume…
