Domain: Model Failure Modes

9 terms

Catastrophic Forgetting Intermediate

Loss of old knowledge when learning new tasks.
Distribution Shift Intermediate

Train/test environment mismatch.
Exposure Bias Intermediate

Differences between training and inference conditions.
Feedback Loop Collapse Intermediate

Model trained on its own outputs degrades quality.
Hallucination Intermediate

Model-generated content that is fluent but unsupported by evidence or incorrect; mitigated by grounding and verification.
Overconfidence Intermediate

Probabilities do not reflect true correctness.
Overgeneralization Intermediate

Applying learned patterns incorrectly.
Prompt Sensitivity Intermediate

Small prompt changes cause large output changes.
Spurious Correlation Intermediate

Model relies on irrelevant signals.