The 60-second answer
SFT teaches desired behavior from demonstrations; preference optimization aligns outputs using comparative feedback. Build high-quality, diverse, policy-consistent supervision before scaling volume.
Build the answer in this order
1
Give the core idea
SFT teaches desired behavior from demonstrations; preference optimization aligns outputs using comparative feedback.
2
Explain how it works
Build high-quality, diverse, policy-consistent supervision before scaling volume.
3
Compare alternatives
Evaluate both target behavior and regressions on broad capabilities and safety.
4
State failure modes + validation
Keep reward/preference data separate from final evaluation data.
A useful interview mental model
This is the shape of a strong answer—not a script to memorize.
01Definition
02Mechanism
03Trade-offs
04Failure modes
05When to use
Senior-level signal
- Discuss reward hacking and preference-model distribution shift.
- Treat alignment as an iterative data-and-evaluation system, not one training stage.
What the interviewer is really testing
Mechanistic understanding, assumptions, trade-offs, and whether you can turn a definition into a model decision.
Likely follow-up questions
What assumption makes this approach work?
When would you choose the strongest alternative instead?
What production or data failure mode changes your answer?
Common weak-answer patterns
- Reciting a definition without mechanism or assumptions.
- Claiming one technique is always better without a data regime.
- Stopping before failure modes, validation, or deployment implications.