Domain: Model Failure Modes

Loss of old knowledge when learning new tasks.

Train/test environment mismatch.

Differences between training and inference conditions.

Model trained on its own outputs degrades quality.

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

Probabilities do not reflect true correctness.

Applying learned patterns incorrectly.

Small prompt changes cause large output changes.

Model relies on irrelevant signals.