🤖 AI Evaluation Engine

Smart QnA uses Vision Transformers and LLMs to evaluate student answers.

👁 Vision Model (ViT)

• Segments answer sheets into visual blocks
• Detects handwritten and diagram components

💬 LLM Matching

• Compares golden answer parts to each student’s response
• Uses similarity scoring to match meaning (not just appearance)

⚙️ Scoring Logic

• If a match is found: assign full/partial marks
• If missing or incorrect: assign 0 or penalize based on rubric rules