You suspect prompt leakage from the eval set into system tuning. How would you respond?
Instruction: Describe how you would handle suspected contamination between the eval suite and the system being optimized.
Context: Tests how the candidate diagnoses the problem, chooses the safest next step, and reasons through recovery. Describe how you would handle suspected contamination between the eval suite and the system being optimized.
Updated
Official answer available
Read the opening below, then unlock the full answer and practical guidance.
I would take the suspicion seriously because leakage can make the benchmark look more predictive than it really is...
Your preparation path
Work through these questions in order. Read the answer aloud, then explain it in your own words.
1. Start with the foundations
Build the vocabulary and explain the core decisions.
2. Apply it to a real workflow
Practice diagnosis, validation, and everyday tradeoffs.
- What signals matter most when evaluating tool-using agents? Free sample
- Your tool-using agent passes answer-level evals and still makes bad tool calls. How would you fix the blind spot? Member answer
- Production feedback says answers feel less trustworthy even though offline metrics are flat. How would you investigate? Member answer
3. Prepare for senior discussions
Explain failure boundaries, recovery, and production choices.
- Human reviewers are expensive, but model graders miss subtle safety issues. How would you balance the two? Member answer
- Two coding agents have different scores and sandbox limits. How would you compare them fairly? Member answer
- An answer tells your LLM grader to award full marks. How would you keep the evaluation trustworthy? Member answer
Related Questions
-
easy
-
easy
-
easy
-
easy
-
easy
-
easy