Databricks-GenAI-Assoc Evaluation and Monitoring Practice Question
A GenAI team at a retail bank runs a RAG assistant on a Databricks Mosaic AI Model Serving endpoint. During a pilot, they captured end-user thumbs-up/down feedback in a Delta table but never joined it to the trace payloads. Six weeks later, hallucination complaints spike, yet the aggregate thumbs-down rate is unchanged. Which approach best resolves this discrepancy using Databricks-native tooling?
⚠ Common exam trap
The trap here is treating an unchanged aggregate feedback rate as proof the application is stable, when the real issue is that feedback was never correlated with trace attributes so segment-level regressions stay invisible.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Join the Delta feedback table to MLflow traces on request_id and segment quality metrics by user cohort, prompt template version, and retrieved document source.
Aggregate feedback rates are diluted by cohort mix, so an unchanged overall thumbs-down percentage can coexist with a severe regression inside one prompt template version, document source, or user cohort. Correlating the stored human feedback with MLflow traces on request_id and then slicing by those dimensions surfaces where the hallucination spike actually lives, which is the prerequisite for any targeted fix.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Increase the endpoint's provisioned concurrency and enable autoscaling so responses are returned faster during peak hours.
Why it's wrong here
Throughput and latency tuning does not affect answer correctness or how feedback is analyzed. The reported symptom is unchanged thumbs-down rates despite rising hallucination complaints, which is an observability/segmentation problem, not a capacity problem. More replicas simply serve the same regressed answers faster.
- ✗
Switch the judge from a smaller open model to a larger frontier model and re-run the nightly evaluation job over the last 30 days.
Why it's wrong here
Upgrading the LLM judge may sharpen automated scoring, but the pilot's signal of record is explicit human thumbs feedback already stored in Delta. Re-scoring historical traces does not link that human feedback to trace attributes, so the team still cannot explain why aggregate sentiment stayed flat while complaints rose.
- ✗
Retrain the embedding model on the latest product documentation and redeploy the index to the Vector Search endpoint.
Why it's wrong here
Re-embedding the corpus addresses retrieval recall, but the scenario gives no evidence that retrieval degraded. The complaint is that aggregate feedback masks a regression; refreshing embeddings neither joins feedback to traces nor segments the metrics, so the team would still be unable to localize the spike to a template or source.
- ✓
Join the Delta feedback table to MLflow traces on request_id and segment quality metrics by user cohort, prompt template version, and retrieved document source.
Why this is correct
Joining feedback rows to MLflow traces via the shared request_id lets the team slice the unchanged aggregate rate by the dimensions that actually moved, exposing whether the spike is confined to one prompt template version or document source. Aggregates hide this because a large silent-majority cohort with stable thumbs-up dilutes the regressed segment.
Visual reference
About these practice questions
This Databricks-GenAI-Assoc question is part of Courseiva's 330-question bank — original exam-style content with full explanations and wrong-answer analysis, never real exam questions or exam dumps. Learn why practice questions differ from exam dumps →
JA
Written and reviewed by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
Last reviewed September 2026 · checked against the official Databricks exam blueprint
This Databricks-GenAI-Assoc practice question is part of Courseiva's free Databricks certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the Databricks-GenAI-Assoc exam.