Year 6 · 35 minutes · Live trained model
Read two fictional facts, then inspect all the words a tiny trained model could choose. Change the facts or the temperature and see what happens.
Student paper journal · Interactive notebook · Editable Python
Try the starting controls, save two runs and reopen a downloaded project. Print the student journal before class. If Python cannot load, use the supplied example below for a prediction and evidence critique; rerunning changed inputs requires the live or native notebook. Do not present paper discussion as a new model experiment. Draw, point or explain aloud. A partner or adult may record your words.
Start with 5–8 minutes of prediction, use about half the lesson for paired comparisons, then reserve at least 10 minutes for a counterexample and an individual explanation. Read the facts aloud before showing the chart. Explain that temperature rearranges probability, not the evidence. Count all vocabulary words, including low-probability ones.
Settings: {'case': 0, 'temperature': 1, 'authored': ''}
Prompt: “pip has red . wattle has blue . pip has …”. The written facts support red. Base model chooses red. Temperature changes probabilities; it does not check the facts.
| word | probability | logit |
|---|---|---|
| pip | 1.3e-05 | -1.35402 |
| wattle | 1e-05 | -1.58286 |
| kiki | 2.1e-05 | -0.89241 |
| bo | 8e-06 | -1.85782 |
| red | 0.999122 | 9.88433 |
| blue | 0.000798 | 2.75241 |
| green | 2e-06 | -3.18138 |
| gold | 2e-06 | -3.0647 |
The table shows up to eight rows. Inspect the notebook for all values, controls and denominators.
Method and scope: probability(word) = softmax(logit / temperature). All 10 vocabulary words are shown; total probability is 1.
Assess four criteria, each 0–2: testable prediction; comparison identifying what stayed fixed; accurate use of exact evidence; counterexample and limited conclusion. 0 means absent or contradicted by the record, 1 means partly supported, and 2 means clear and supported. A surprising result earns no penalty.
Beginning response: ‘The picture looks right.’ Ask for a particular case. Developing response: names one number without its setting. Ask what comparison supports it. Secure response: states the setting and measured change and separates the notebook's result from the source paper's claim. Extending response: authors or reserves a new case, tests an alternative explanation, and revises the claim if needed. These are marking examples, not pupil data.
Extension: Draw two probability bars on paper. Older learners can check that all probabilities sum to one.
System and data: Llama-3-8B-Instruct and DeepSeek-R1-Distill-Llama-8B on tinyMMLU
Method: Resamples continuations at shared prefixes and compares uncertainty estimates across sample budgets, spacing and smoothing.
Task: Try the same starting clue many times and count different endings. Use the linked notebook to make two observations. Draw or describe one result and one thing this activity cannot tell us about the source system.
Boundary: Our finite branching simulator is not a reasoning LLM. More samples reduce sampling noise, not model bias.
System and data: Paired language-model checkpoints and controlled fine-tunes
Method: Amplifies output-score differences to expose changed behaviour; increased discovery frequency is not natural failure prevalence.
Task: Compare two versions on the same question. Use the linked notebook to make two observations. Draw or describe one result and one thing this activity cannot tell us about the source system.
Boundary: Amplified discovery is not an estimate of how often the original model fails.
Independent Brightlab adaptations; no Goodfire endorsement or full reproduction claim. Source methods reviewed 7 September 2026. Curriculum connections are selected planning links; confirm your school syllabus.