Evaluation & Trust

Build test sets that tell the truth, use LLM judges without being fooled by them, measure hallucination and bias, evaluate multi-turn agents, and turn all of it into a trust score you can ship on.

Lessons

1. 1. Test sets that tell the truth

   An evaluation is only as honest as its test set. Start small, draw from real failures, and keep it away from training.

   8 min

2. 2. LLM-as-a-judge, and its biases

   Strong models grade open-ended answers about as consistently as people do. They also prefer longer answers, the first answer, and their own writing.

   9 min

3. 3. Measuring hallucination and bias

   Two failure modes that erode trust fastest. Both can be measured, as long as you break them into checkable parts.

   9 min

4. 4. Evaluating agents: trials, transcripts and pass^k

   Agents act over many turns and change the world around them. Grade the outcome, read the transcript, and run every task more than once.

   10 min

5. 5. Latency, trust scores and the ship decision

   Users feel the slowest responses, not the average. Combine quality and speed into one decision you can defend.

   7 min

Sources

Every source cited across this course's lessons.

1. [1]Anthropic (2026). Demystifying evals for AI agents. Anthropic Engineering, January 2026.
2. [2]Zinkevich, M. (2017). Rules of Machine Learning: Best Practices for ML Engineering. Google for Developers.
3. [3]Zheng, L. et al. (2023). Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena. NeurIPS 2023 · arXiv:2306.05685.
4. [4]Shankar, S. et al. (2024). Who Validates the Validators? Aligning LLM-Assisted Evaluation of LLM Outputs with Human Preferences. UIST 2024 · arXiv:2404.12272.
5. [5]Ji, Z. et al. (2023). Survey of Hallucination in Natural Language Generation. ACM Computing Surveys · arXiv:2202.03629.
6. [6]Min, S. et al. (2023). FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generation. EMNLP 2023 · arXiv:2305.14251.
7. [7]Parrish, A. et al. (2022). BBQ: A Hand-Built Bias Benchmark for Question Answering. Findings of ACL 2022 · arXiv:2110.08193.
8. [8]Jimenez, C. E. et al. (2023). SWE-bench: Can Language Models Resolve Real-World GitHub Issues?. ICLR 2024 · arXiv:2310.06770.
9. [9]Yao, S. et al. (2024). τ-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains. arXiv:2406.12045.
10. [10]Dean, J. & Barroso, L. A. (2013). The Tail at Scale. Communications of the ACM 56(2).
11. [11]Beyer, B. et al. (eds.) (2016). Site Reliability Engineering, ch. 6: Monitoring Distributed Systems. O'Reilly / Google.
12. [12]Mitchell, M. et al. (2019). Model Cards for Model Reporting. FAT* 2019 · arXiv:1810.03993.