What does conflict-driven clause learning add to plain backtracking?

• A guarantee of polynomial worst case
• Faster variable ordering only
• A learned clause per failure that prunes repeats

What does a satisfying assignment mean for the encoded problem?

• One rule may be ignored
• Every rule is obeyed at once
• The search ran out of time

A solver that is fast on factory scheduling instances disproves NP-completeness.

Circle one:   True   False

A solver keeps failing the same region in different branches. What went wrong?

• It learned nothing from the failures
• It learned too many clauses
• It encoded too few variables

A and B cannot run at the same time. Which clause encodes that?

• A or B
• Not-A or not-B
• A and B

An assignment satisfies 99 of 100 clauses. Is the formula satisfied?

• Yes, nearly all hold
• Only if the last clause is redundant
• No, every clause must hold

A new rule says task C needs task A finished first. With A-first meaning A, which clause fits C-before-A being forbidden?

• Not-C-first or A-first
• C-first and A-first
• Not-C-first and not-A-first

Ned says clause learning improves the worst-case complexity class. What is wrong?

• Learning only slows solvers down
• The worst case is unchanged; practice is what improves
• Backtracking already learns every clause