Lesson overview | Previous part | Lesson overview

Experiment Tracking and Reproducibility: Part 7: LLM and GenAI Tracking to References

7. LLM and GenAI Tracking

LLM and GenAI Tracking develops the part of experiment tracking and reproducibility assigned by the approved Chapter 19 table of contents. The treatment is production-focused: every idea is connected to a versioned artifact, measurable signal, release decision, or incident response.

7.1 prompt versions

Prompt versions is part of the canonical scope of Experiment Tracking and Reproducibility. In production ML, the useful question is not only whether the model can be trained, but whether the surrounding artifact, signal, or control can be named, versioned, measured, and recovered after a failure.

For this section, the working object is run metadata, reproducibility envelopes, model registries, statistical comparison, and production promotion gates. The notation below treats production systems as mathematical objects because that is how incidents become diagnosable. A dataset, feature, run, trace, or endpoint that lacks a stable identifier cannot be compared across time.

$$E(r) = \{\mathcal{D}_v, c, e, s, H\}.$$

The formula is intentionally simple. It says that prompt versions should be reduced to a measurable object before anyone argues about dashboards or tools. Once the object is measurable, the system can decide whether to accept, warn, rollback, retrain, or escalate.

Examples of prompt versions in a real system:

1. A production pipeline records the input version, transformation code hash, model version, and endpoint version before serving predictions.
2. An LLM application logs prompt version, retrieval index version, tool span, latency, token count, and guardrail action for each trace.
3. A release gate compares the candidate model against the current model on quality, safety, latency, and cost before promotion.

Non-examples that often look similar but fail the production contract:

1. A manually named file like final_dataset.csv with no hash, schema, lineage, or owner.
2. A metric screenshot pasted into chat without the run id, evaluation dataset, seed, or model artifact.
3. A dashboard alert with no threshold rationale, no escalation rule, and no rollback candidate.

The AI connection is concrete. Modern ML and LLM systems are compound systems: data pipelines, feature stores, model registries, inference servers, retrievers, tools, evaluators, and safety layers. Prompt versions is one place where the compound system either becomes observable or becomes technical debt.

Operational checklist for prompt versions:

• State the artifact or signal being controlled.
• Give it a stable id and version.
• Define the metric or predicate that decides whether it is valid.
• Log the dependency chain needed to reproduce it.
• Attach an owner and a response action.
• Test the check in continuous integration or release gating.

A useful mental model is to treat every production ML component as a function with preconditions and postconditions. If $u$ is the upstream artifact and $z$ is the downstream artifact, the production question is whether the relation $u \mapsto z$ can be replayed and audited.

$$z = T(u; c, e),$$

where $T$ is the transformation, $c$ is code or configuration, and $e$ is the execution environment. The hidden technical debt appears when any of $u$, $c$, or $e$ is missing from the record.

In notebooks, this subsection will be represented with small synthetic arrays, graphs, traces, or counters rather than external services. The point is not to mimic a vendor tool. The point is to make the mathematics of prompt versions executable enough to test.

Boundary note: this chapter assumes the evaluation methods from Chapter 17, the safety policy ideas from Chapter 18, and the data documentation work from Chapter 16. Here we focus on the production machinery that makes those ideas run repeatedly.

Failure analysis for prompt versions should be written before the incident occurs. A good production note asks what can be stale, missing, corrupted, delayed, unaudited, or too expensive. Each answer should correspond to one observable signal and one response action.

The production design pattern is therefore not just to calculate a value. It is to calculate a value, compare it with a declared rule, log the evidence, and make the next action unambiguous. That four-step pattern will reappear across all Chapter 19 notebooks.

7.2 evaluation traces

Evaluation traces is part of the canonical scope of Experiment Tracking and Reproducibility. In production ML, the useful question is not only whether the model can be trained, but whether the surrounding artifact, signal, or control can be named, versioned, measured, and recovered after a failure.

For this section, the working object is run metadata, reproducibility envelopes, model registries, statistical comparison, and production promotion gates. The notation below treats production systems as mathematical objects because that is how incidents become diagnosable. A dataset, feature, run, trace, or endpoint that lacks a stable identifier cannot be compared across time.

$$\Delta = m(r_a)-m(r_b).$$

The formula is intentionally simple. It says that evaluation traces should be reduced to a measurable object before anyone argues about dashboards or tools. Once the object is measurable, the system can decide whether to accept, warn, rollback, retrain, or escalate.

Examples of evaluation traces in a real system:

1. A production pipeline records the input version, transformation code hash, model version, and endpoint version before serving predictions.
2. An LLM application logs prompt version, retrieval index version, tool span, latency, token count, and guardrail action for each trace.
3. A release gate compares the candidate model against the current model on quality, safety, latency, and cost before promotion.

Non-examples that often look similar but fail the production contract:

1. A manually named file like final_dataset.csv with no hash, schema, lineage, or owner.
2. A metric screenshot pasted into chat without the run id, evaluation dataset, seed, or model artifact.
3. A dashboard alert with no threshold rationale, no escalation rule, and no rollback candidate.

The AI connection is concrete. Modern ML and LLM systems are compound systems: data pipelines, feature stores, model registries, inference servers, retrievers, tools, evaluators, and safety layers. Evaluation traces is one place where the compound system either becomes observable or becomes technical debt.

Operational checklist for evaluation traces:

• State the artifact or signal being controlled.
• Give it a stable id and version.
• Define the metric or predicate that decides whether it is valid.
• Log the dependency chain needed to reproduce it.
• Attach an owner and a response action.
• Test the check in continuous integration or release gating.

A useful mental model is to treat every production ML component as a function with preconditions and postconditions. If $u$ is the upstream artifact and $z$ is the downstream artifact, the production question is whether the relation $u \mapsto z$ can be replayed and audited.

$$z = T(u; c, e),$$

where $T$ is the transformation, $c$ is code or configuration, and $e$ is the execution environment. The hidden technical debt appears when any of $u$, $c$, or $e$ is missing from the record.

In notebooks, this subsection will be represented with small synthetic arrays, graphs, traces, or counters rather than external services. The point is not to mimic a vendor tool. The point is to make the mathematics of evaluation traces executable enough to test.

Boundary note: this chapter assumes the evaluation methods from Chapter 17, the safety policy ideas from Chapter 18, and the data documentation work from Chapter 16. Here we focus on the production machinery that makes those ideas run repeatedly.

Failure analysis for evaluation traces should be written before the incident occurs. A good production note asks what can be stale, missing, corrupted, delayed, unaudited, or too expensive. Each answer should correspond to one observable signal and one response action.

The production design pattern is therefore not just to calculate a value. It is to calculate a value, compare it with a declared rule, log the evidence, and make the next action unambiguous. That four-step pattern will reappear across all Chapter 19 notebooks.

7.3 dataset model prompt triples

Dataset model prompt triples is part of the canonical scope of Experiment Tracking and Reproducibility. In production ML, the useful question is not only whether the model can be trained, but whether the surrounding artifact, signal, or control can be named, versioned, measured, and recovered after a failure.

For this section, the working object is run metadata, reproducibility envelopes, model registries, statistical comparison, and production promotion gates. The notation below treats production systems as mathematical objects because that is how incidents become diagnosable. A dataset, feature, run, trace, or endpoint that lacks a stable identifier cannot be compared across time.

$$r = (\boldsymbol{\lambda}, s, e, \mathcal{D}_v, c, \mathbf{m}, \mathcal{A}_r).$$

The formula is intentionally simple. It says that dataset model prompt triples should be reduced to a measurable object before anyone argues about dashboards or tools. Once the object is measurable, the system can decide whether to accept, warn, rollback, retrain, or escalate.

Examples of dataset model prompt triples in a real system:

1. A production pipeline records the input version, transformation code hash, model version, and endpoint version before serving predictions.
2. An LLM application logs prompt version, retrieval index version, tool span, latency, token count, and guardrail action for each trace.
3. A release gate compares the candidate model against the current model on quality, safety, latency, and cost before promotion.

Non-examples that often look similar but fail the production contract:

1. A manually named file like final_dataset.csv with no hash, schema, lineage, or owner.
2. A metric screenshot pasted into chat without the run id, evaluation dataset, seed, or model artifact.
3. A dashboard alert with no threshold rationale, no escalation rule, and no rollback candidate.

The AI connection is concrete. Modern ML and LLM systems are compound systems: data pipelines, feature stores, model registries, inference servers, retrievers, tools, evaluators, and safety layers. Dataset model prompt triples is one place where the compound system either becomes observable or becomes technical debt.

Operational checklist for dataset model prompt triples:

• State the artifact or signal being controlled.
• Give it a stable id and version.
• Define the metric or predicate that decides whether it is valid.
• Log the dependency chain needed to reproduce it.
• Attach an owner and a response action.
• Test the check in continuous integration or release gating.

A useful mental model is to treat every production ML component as a function with preconditions and postconditions. If $u$ is the upstream artifact and $z$ is the downstream artifact, the production question is whether the relation $u \mapsto z$ can be replayed and audited.

$$z = T(u; c, e),$$

where $T$ is the transformation, $c$ is code or configuration, and $e$ is the execution environment. The hidden technical debt appears when any of $u$, $c$, or $e$ is missing from the record.

In notebooks, this subsection will be represented with small synthetic arrays, graphs, traces, or counters rather than external services. The point is not to mimic a vendor tool. The point is to make the mathematics of dataset model prompt triples executable enough to test.

Boundary note: this chapter assumes the evaluation methods from Chapter 17, the safety policy ideas from Chapter 18, and the data documentation work from Chapter 16. Here we focus on the production machinery that makes those ideas run repeatedly.

Failure analysis for dataset model prompt triples should be written before the incident occurs. A good production note asks what can be stale, missing, corrupted, delayed, unaudited, or too expensive. Each answer should correspond to one observable signal and one response action.

The production design pattern is therefore not just to calculate a value. It is to calculate a value, compare it with a declared rule, log the evidence, and make the next action unambiguous. That four-step pattern will reappear across all Chapter 19 notebooks.

7.4 token and cost metrics

Token and cost metrics is part of the canonical scope of Experiment Tracking and Reproducibility. In production ML, the useful question is not only whether the model can be trained, but whether the surrounding artifact, signal, or control can be named, versioned, measured, and recovered after a failure.

For this section, the working object is run metadata, reproducibility envelopes, model registries, statistical comparison, and production promotion gates. The notation below treats production systems as mathematical objects because that is how incidents become diagnosable. A dataset, feature, run, trace, or endpoint that lacks a stable identifier cannot be compared across time.

$$\bar{m} = \frac{1}{K}\sum_{k=1}^{K}m_k, \qquad \widehat{\operatorname{SE}}(\bar{m}) = \frac{s_m}{\sqrt{K}}.$$

The formula is intentionally simple. It says that token and cost metrics should be reduced to a measurable object before anyone argues about dashboards or tools. Once the object is measurable, the system can decide whether to accept, warn, rollback, retrain, or escalate.

Examples of token and cost metrics in a real system:

1. A production pipeline records the input version, transformation code hash, model version, and endpoint version before serving predictions.
2. An LLM application logs prompt version, retrieval index version, tool span, latency, token count, and guardrail action for each trace.
3. A release gate compares the candidate model against the current model on quality, safety, latency, and cost before promotion.

Non-examples that often look similar but fail the production contract:

1. A manually named file like final_dataset.csv with no hash, schema, lineage, or owner.
2. A metric screenshot pasted into chat without the run id, evaluation dataset, seed, or model artifact.
3. A dashboard alert with no threshold rationale, no escalation rule, and no rollback candidate.

The AI connection is concrete. Modern ML and LLM systems are compound systems: data pipelines, feature stores, model registries, inference servers, retrievers, tools, evaluators, and safety layers. Token and cost metrics is one place where the compound system either becomes observable or becomes technical debt.

Operational checklist for token and cost metrics:

• State the artifact or signal being controlled.
• Give it a stable id and version.
• Define the metric or predicate that decides whether it is valid.
• Log the dependency chain needed to reproduce it.
• Attach an owner and a response action.
• Test the check in continuous integration or release gating.

A useful mental model is to treat every production ML component as a function with preconditions and postconditions. If $u$ is the upstream artifact and $z$ is the downstream artifact, the production question is whether the relation $u \mapsto z$ can be replayed and audited.

$$z = T(u; c, e),$$

where $T$ is the transformation, $c$ is code or configuration, and $e$ is the execution environment. The hidden technical debt appears when any of $u$, $c$, or $e$ is missing from the record.

In notebooks, this subsection will be represented with small synthetic arrays, graphs, traces, or counters rather than external services. The point is not to mimic a vendor tool. The point is to make the mathematics of token and cost metrics executable enough to test.

Boundary note: this chapter assumes the evaluation methods from Chapter 17, the safety policy ideas from Chapter 18, and the data documentation work from Chapter 16. Here we focus on the production machinery that makes those ideas run repeatedly.

Failure analysis for token and cost metrics should be written before the incident occurs. A good production note asks what can be stale, missing, corrupted, delayed, unaudited, or too expensive. Each answer should correspond to one observable signal and one response action.

The production design pattern is therefore not just to calculate a value. It is to calculate a value, compare it with a declared rule, log the evidence, and make the next action unambiguous. That four-step pattern will reappear across all Chapter 19 notebooks.

7.5 judge-version tracking

Judge-version tracking is part of the canonical scope of Experiment Tracking and Reproducibility. In production ML, the useful question is not only whether the model can be trained, but whether the surrounding artifact, signal, or control can be named, versioned, measured, and recovered after a failure.

For this section, the working object is run metadata, reproducibility envelopes, model registries, statistical comparison, and production promotion gates. The notation below treats production systems as mathematical objects because that is how incidents become diagnosable. A dataset, feature, run, trace, or endpoint that lacks a stable identifier cannot be compared across time.

$$E(r) = \{\mathcal{D}_v, c, e, s, H\}.$$

The formula is intentionally simple. It says that judge-version tracking should be reduced to a measurable object before anyone argues about dashboards or tools. Once the object is measurable, the system can decide whether to accept, warn, rollback, retrain, or escalate.

Examples of judge-version tracking in a real system:

1. A production pipeline records the input version, transformation code hash, model version, and endpoint version before serving predictions.
2. An LLM application logs prompt version, retrieval index version, tool span, latency, token count, and guardrail action for each trace.
3. A release gate compares the candidate model against the current model on quality, safety, latency, and cost before promotion.

Non-examples that often look similar but fail the production contract:

1. A manually named file like final_dataset.csv with no hash, schema, lineage, or owner.
2. A metric screenshot pasted into chat without the run id, evaluation dataset, seed, or model artifact.
3. A dashboard alert with no threshold rationale, no escalation rule, and no rollback candidate.

The AI connection is concrete. Modern ML and LLM systems are compound systems: data pipelines, feature stores, model registries, inference servers, retrievers, tools, evaluators, and safety layers. Judge-version tracking is one place where the compound system either becomes observable or becomes technical debt.

Operational checklist for judge-version tracking:

• State the artifact or signal being controlled.
• Give it a stable id and version.
• Define the metric or predicate that decides whether it is valid.
• Log the dependency chain needed to reproduce it.
• Attach an owner and a response action.
• Test the check in continuous integration or release gating.

A useful mental model is to treat every production ML component as a function with preconditions and postconditions. If $u$ is the upstream artifact and $z$ is the downstream artifact, the production question is whether the relation $u \mapsto z$ can be replayed and audited.

$$z = T(u; c, e),$$

where $T$ is the transformation, $c$ is code or configuration, and $e$ is the execution environment. The hidden technical debt appears when any of $u$, $c$, or $e$ is missing from the record.

In notebooks, this subsection will be represented with small synthetic arrays, graphs, traces, or counters rather than external services. The point is not to mimic a vendor tool. The point is to make the mathematics of judge-version tracking executable enough to test.

Boundary note: this chapter assumes the evaluation methods from Chapter 17, the safety policy ideas from Chapter 18, and the data documentation work from Chapter 16. Here we focus on the production machinery that makes those ideas run repeatedly.

Failure analysis for judge-version tracking should be written before the incident occurs. A good production note asks what can be stale, missing, corrupted, delayed, unaudited, or too expensive. Each answer should correspond to one observable signal and one response action.

The production design pattern is therefore not just to calculate a value. It is to calculate a value, compare it with a declared rule, log the evidence, and make the next action unambiguous. That four-step pattern will reappear across all Chapter 19 notebooks.

8. Common Mistakes

9. Exercises

1. (*) Design a production ML check related to experiment tracking and reproducibility.

   • (a) Define the object being checked using mathematical notation.
   • (b) State the metric, predicate, or threshold used to decide pass/fail.
   • (c) Explain which artifact versions must be logged.
   • (d) Give one failure case and one rollback or escalation action.

2. (*) Design a production ML check related to experiment tracking and reproducibility.

   • (a) Define the object being checked using mathematical notation.
   • (b) State the metric, predicate, or threshold used to decide pass/fail.
   • (c) Explain which artifact versions must be logged.
   • (d) Give one failure case and one rollback or escalation action.

3. (*) Design a production ML check related to experiment tracking and reproducibility.

   • (a) Define the object being checked using mathematical notation.
   • (b) State the metric, predicate, or threshold used to decide pass/fail.
   • (c) Explain which artifact versions must be logged.
   • (d) Give one failure case and one rollback or escalation action.

4. (**) Design a production ML check related to experiment tracking and reproducibility.

   • (a) Define the object being checked using mathematical notation.
   • (b) State the metric, predicate, or threshold used to decide pass/fail.
   • (c) Explain which artifact versions must be logged.
   • (d) Give one failure case and one rollback or escalation action.

5. (**) Design a production ML check related to experiment tracking and reproducibility.

   • (a) Define the object being checked using mathematical notation.
   • (b) State the metric, predicate, or threshold used to decide pass/fail.
   • (c) Explain which artifact versions must be logged.
   • (d) Give one failure case and one rollback or escalation action.

6. (**) Design a production ML check related to experiment tracking and reproducibility.

   • (a) Define the object being checked using mathematical notation.
   • (b) State the metric, predicate, or threshold used to decide pass/fail.
   • (c) Explain which artifact versions must be logged.
   • (d) Give one failure case and one rollback or escalation action.

7. (***) Design a production ML check related to experiment tracking and reproducibility.

   • (a) Define the object being checked using mathematical notation.
   • (b) State the metric, predicate, or threshold used to decide pass/fail.
   • (c) Explain which artifact versions must be logged.
   • (d) Give one failure case and one rollback or escalation action.

8. (***) Design a production ML check related to experiment tracking and reproducibility.

   • (a) Define the object being checked using mathematical notation.
   • (b) State the metric, predicate, or threshold used to decide pass/fail.
   • (c) Explain which artifact versions must be logged.
   • (d) Give one failure case and one rollback or escalation action.

9. (***) Design a production ML check related to experiment tracking and reproducibility.

   • (a) Define the object being checked using mathematical notation.
   • (b) State the metric, predicate, or threshold used to decide pass/fail.
   • (c) Explain which artifact versions must be logged.
   • (d) Give one failure case and one rollback or escalation action.

10. (***) Design a production ML check related to experiment tracking and reproducibility.

• (a) Define the object being checked using mathematical notation.
• (b) State the metric, predicate, or threshold used to decide pass/fail.
• (c) Explain which artifact versions must be logged.
• (d) Give one failure case and one rollback or escalation action.

10. Why This Matters for AI

11. Conceptual Bridge

Experiment Tracking and Reproducibility sits after the chapters on data construction, evaluation, and alignment because production systems combine all three. Chapter 16 explains how reliable datasets are assembled. Chapter 17 explains how models are measured. Chapter 18 explains how desired behavior and safety constraints are specified. Chapter 19 asks whether those ideas survive contact with changing data, users, services, and costs.

The backward bridge is operational memory. If a model fails today, the team must recover the data, code, environment, model, endpoint, prompt, retriever, guardrail, and metric definitions that produced the behavior. That is why the notation in this chapter emphasizes hashes, graphs, traces, thresholds, and predicates.

The forward bridge is broader mathematical maturity. Later chapters return to signal processing, learning theory, causal inference, game theory, measure theory, and geometry. Production ML uses those ideas under constraints: bounded latency, incomplete labels, shifting distributions, and costly human attention.

+--------------------------------------------------------------+
| Chapter 16: data construction and governance                 |
| Chapter 17: evaluation and reliability                       |
| Chapter 18: alignment and safety                             |
| Chapter 19: production ML and MLOps                          |
|   artifact -> endpoint -> telemetry -> alert -> retrain      |
| Chapter 20+: mathematical tools for deeper modeling          |
+--------------------------------------------------------------+

References

• MLflow. Model registry tutorial and experiment tracking documentation. https://www.mlflow.org/docs/latest/ml/model-registry/tutorial
• Kubeflow. Pipelines concepts. https://www.kubeflow.org/docs/components/pipelines/concepts/pipeline/
• Mitchell et al.. Model Cards for Model Reporting. https://arxiv.org/abs/1810.03993
• Google Cloud. MLOps continuous delivery for ML. https://cloud.google.com/solutions/machine-learning/mlops-continuous-delivery-and-automation-pipelines-in-machine-learning