Prompt context and fine tuning

Your goal

Choose whether to change an instruction, supply current facts or investigate further training.

To be able to choose a suitable way to steer a model and explain whether its learned settings change.

Before you start

A menu changes today. Which gives current information?

Choices: Today’s approved menu / A more confident sentence

Words for this lesson

Prompt

Instructions or information given for a response.

Context

Information available while the response is produced.

Fine tuning

Further training that adjusts learned settings.

The idea

A prompt can change instructions such as reply in two sentences. Context can supply current facts such as a new catalogue. These input changes do not retrain the model in this activity. Fine tuning is further training that adjusts learned settings. Choose the change that fits the problem, then test the result. Some tasks need ordinary software or a person.

A completed example

Practice: the source says a club meets on Tuesday. Asking for one short sentence changes format. Supplying a new timetable can change the available fact. Neither input action by itself trains new weights.

Check your understanding

Adding a current source changes ...

1. Available context
2. All learned weights automatically
3. Every future source

Fine tuning involves ...

1. Further training
2. Only a font change
3. Only copying a timetable

A date comparison might be best handled by ...

1. An ordinary programmed rule
2. Always a large language model
3. Private-data fine tuning

Task 1 · Make a prediction

Will adding a current timetable retrain weights

Options: Yes / No / Not sure

Write your prediction, or state that you skipped it.

Try it using supplied records

1. Read the same fictional question with no source, then with the current source. Predict what information is available.
2. Press Show scripted reply. Compare the reply with the source.
3. Change only the format instruction to two short sentences. The fixed-version label remains V1.
4. Open a proposed fine-tuning card. Predict whether learned settings would change if developers carried it out. No training is performed here.

Authored illustrations · no training performed

Complete data and source records

Expected results · supplied

These are supplied reference results. They are separate from your own explanation. No code needs to run to read them.

1. Board outputs are hard-coded authored illustrations, labelled before reveal. Views 1-3 keep model version V1 and do not train. The fine-tuning card says Proposed further training, untested; do not manufacture a successful V2 result.

2. A clearer prompt, no weight change; B approved current context or retrieval, no weight change required; C consider fine tuning with suitable data and independent evaluation, settings would change if performed; D ordinary date comparison, no trained model needed; E pause for an authorised human process.

3. Allow a reasoned alternative first action. The fixed-choice answer is suggested, while the independent explanation remains teacher assessed. Learned settings field supports No, Would change if trained, and No model needed.

Help with this step

Task 2 · Use the evidence

1. For each A to E school brief, choose a sensible first action.
2. State whether learned model settings change and name a test or limit.
3. Explain the changing-catalogue choice using source freshness and one check, in two or three sentences.

My explanation

Write in your school assignment. Keep these answers in your own words.

Case A

• First action: Clearer prompt / Current approved context or retrieval / Consider fine tuning with evaluation / Ordinary software / Pause for human review
• Learned settings scope: No change required / Would change if trained / No trained model needed
• Test or limit

Case B

• First action: Clearer prompt / Current approved context or retrieval / Consider fine tuning with evaluation / Ordinary software / Pause for human review
• Learned settings scope: No change required / Would change if trained / No trained model needed
• Test or limit

Case C

• First action: Clearer prompt / Current approved context or retrieval / Consider fine tuning with evaluation / Ordinary software / Pause for human review
• Learned settings scope: No change required / Would change if trained / No trained model needed
• Test or limit

Case D

• First action: Clearer prompt / Current approved context or retrieval / Consider fine tuning with evaluation / Ordinary software / Pause for human review
• Learned settings scope: No change required / Would change if trained / No trained model needed
• Test or limit

Case E

• First action: Clearer prompt / Current approved context or retrieval / Consider fine tuning with evaluation / Ordinary software / Pause for human review
• Learned settings scope: No change required / Would change if trained / No trained model needed
• Test or limit

Your explanation

• Why the catalogue needs this approach

Explain a fresh example

The library changes its catalogue every week. Recommend a first approach, say whether learned settings must change, and name a test.

My confidence

I need another explanation / I can explain with help / I can explain with my own example

My next step

Write what you will do next.

Optional depth

Why might a fine-tuned version still need current sources and prompts? How could developers compare it with an earlier version?

What this activity can show

This site shows authored replies and a proposed training intervention. It does not fine-tune a model. A fine-tuned version is replaceable; further training does not guarantee current facts or safety.

This is a student learning record. Submission is completed through the school assignment system.