Student code explanation assessment on SelfCode balanced 5-fold CV 2.0 (test)
98F1 ScoreLLM-Behavior
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LLM-BehaviorApproach=LLM-based, Threshold tuning=None, Fine-tuning=None2026.05 | 98 | 96 | |
| LLM-NoDef.Approach=LLM-based, Threshold tuning=None, Fine-tuning=None2026.05 | 98 | 96 | |
| LLM-Constr.Approach=LLM-based, Threshold tuning=None, Fine-tuning=None2026.05 | 89 | 80 | |
| GPT-SEApproach=Semantic Similarity, Threshold=Optimal2026.05 | 72 | 65 | |
| RoBERTaApproach=Semantic Similarity, Threshold=Optimal2026.05 | 70 | 61 | |
| Deep TutorApproach=Semantic Similarity, Threshold=Optimal2026.05 | 69 | 60 |