Adversarial Question Generation For Misconception Targeting

via “adversarial unanswerable question generation and validation”

150K reading comprehension questions including unanswerable ones.

Unique: Pioneered adversarial unanswerable questions in QA benchmarks by having crowdworkers explicitly write questions that CANNOT be answered from a passage. This is fundamentally different from randomly sampling unanswerable questions; adversarial construction ensures questions are plausible but genuinely unanswerable.

vs others: More challenging than datasets with random negative examples (e.g., MS MARCO) because adversarial questions require models to understand semantic relevance, not just keyword matching, to distinguish answerable from unanswerable.

via “adversarial-question-generation-for-misconception-targeting”

817 adversarial questions measuring model truthfulness vs misconceptions.

Unique: Explicitly targets common human misconceptions through adversarial question design rather than generic factuality testing; combines truthfulness evaluation (factual correctness) with informativeness scoring (useful detail), addressing both accuracy and utility in a single benchmark framework

vs others: More targeted than generic QA benchmarks (SQuAD, Natural Questions) because it adversarially crafts questions to expose model susceptibility to false beliefs rather than measuring generic reading comprehension or retrieval accuracy

via “adversarial-filtered multiple-choice evaluation”

70K commonsense reasoning questions with adversarial distractors.

Unique: Uses adversarial filtering where distractors are selected based on measured model confusion rather than human-written plausibility, creating a dataset that specifically targets machine weaknesses while maintaining human interpretability. This two-stage LLM-generation + human-validation approach is more scalable than purely human-written distractors while maintaining higher quality than random negatives.

vs others: Harder than SWAG (predecessor) because distractors are adversarially selected for model confusion, and more human-aligned than synthetic reasoning datasets because human accuracy (95.6%) validates that hard-for-models questions remain easy for humans.

via “common-pitfalls-and-misconceptions-clarification”

provides a step-by-step guide for beginners to understand and develop AI skills. It covers foundational topics like programming (Python), mathematics, and machine learning, progressing to advanced concepts such as deep learning and neural networks.

via “assessment-generation-and-question-banking”

Unique: Combines procedural generation (for math/science) with LLM synthesis (for open-ended questions) and maintains question metadata (difficulty, discrimination) to enable adaptive selection rather than random question assignment

vs others: More scalable than manually curated question banks because it generates unlimited questions while maintaining quality through template-based generation and LLM synthesis, reducing teacher workload

via “misconception-detection-and-correction”