Mapping LLMs Unplugged to the Australian Curriculum

This document maps the LLMs Unplugged lessons to the Australian Curriculum v9.0 so that teachers can drop them into existing programs and administrators can see what coverage a unit of LLMs Unplugged lessons delivers.

Status: draft, May 2026. Codes are taken from the official ACARA machine-readable workbook (curriculum-workbook.xlsx, downloads page) and have not yet been reviewed by an external curriculum expert.

Purpose and scope#

LLMs Unplugged teaches the core mechanics of large language models (LLMs) through hands-on, pen-and-paper-and-dice activities. The pedagogy is unplugged (no computers required) but the concepts are the same ones used in production AI systems. This puts the resource in a curriculum sweet spot: it is rigorous enough to count as Digital Technologies content, but its activities are also legible to Mathematics (probability, statistics), English (language and text structure), and the general capabilities (Critical and Creative Thinking, Digital Literacy, Ethical Understanding).

The mapping below is meant to help two audiences:

• classroom teachers, who need to know which lesson aligns with a specific content description they are already required to teach
• curriculum coordinators and administrators, who need to see the cumulative coverage a sequence of LLMs Unplugged lessons gives them across learning areas

State-level syllabi (NSW NESA, Victorian VCAA, QCAA, SCSA, etc.) inherit from the ACARA F–10 v9 codes used here, so this mapping is portable. Senior secondary (Years 11–12) is out of scope.

Mapping principles#

1. Anchor to content description codes, not topic labels. The codes (e.g.
   AC9TDI8P02) are stable and auditable; topic labels are not.
2. Distinguish primary from secondary alignment. A primary mapping means the lesson directly addresses the content description and could be used as a teaching/assessment vehicle for it. A secondary mapping means the lesson touches on the content but should not be relied on as the main learning experience.
3. Year band, not exact year. The materials work best in 5–6, 7–8, and 9–10 bands; we recommend bands rather than single years.
4. Cross-curricular by design. Almost every lesson maps to more than one learning area. We list each alignment under its strongest learning area first.
5. Be honest about what’s out of scope. LLMs Unplugged does not teach programming syntax, formal computational complexity, or the full mathematical apparatus of probability theory. Where a lesson stops short of a content description, we say so.

At-a-glance#

Per-lesson mappings#

Each lesson gets a card with: year band, primary codes, secondary codes, general capabilities, and a one-line justification per code.

Weighted Randomness#

Probability sampling from a target distribution using dice and beads. The mathematical bedrock for everything that follows.

• Best fit: Years 5–8
• Primary
  • AC9M5P02 — conduct repeated chance experiments with equally and not equally likely outcomes; use frequency to compare and estimate likelihood.
  • AC9M6P02 — run simulations with increasing trials; compare observed with expected results.
  • AC9M7P02 — conduct repeated chance experiments and run simulations; compare predictions with observed results.
• Secondary
  • AC9M6P01 — probabilities lie on numerical scales; use fractions, decimals and percentages.
  • AC9TDI6K03 — digital systems represent data using numbers (because dice rolls are the number representation here).
• General capabilities: Numeracy (recognising and using patterns and relationships); Critical and Creative Thinking (Inquiring).

Training#

Build a bigram model by counting which words follow which in a corpus.

• Best fit: Years 5–8
• Primary
  • AC9TDI8P01 — acquire, store and validate data from a range of sources. The corpus is the source; the grid is the store.
  • AC9TDI8P02 — analyse and visualise data to identify patterns. Token co-occurrence is the pattern.
  • AC9TDI6P05 — implement algorithms involving control structures, variables and input (the counting algorithm).
• Secondary
  • AC9M5ST01 / AC9M6ST01 (statistics strand) — categorise and describe data and frequencies.
  • AC9E7LA08 — vocabulary in building specialist knowledge.
• General capabilities: Digital Literacy (Investigating > Locate, generate and access data); Numeracy.

Generation#

Use the trained bigram model to generate new text via weighted sampling.

• Best fit: Years 5–8
• Primary
  • AC9M5P02, AC9M7P01 — sample space and assigning probabilities to outcomes.
  • AC9TDI6P05 — implement algorithms with control structures and input (the generation loop).
• Secondary
  • AC9TDI8P06 — trace algorithms to predict output for a given input.
  • AC9E5LA03 — describe how texts use language features and are organised into stages and phases (when comparing generated and source text).
• General capabilities: Numeracy; Critical and Creative Thinking (Generating).

Pre-trained Model Generation#

Use a pre-built booklet to generate text without training your own model. Introduces the consume-vs-build distinction that underpins most real-world AI usage.

• Best fit: Years 5–8
• Primary
  • AC9TDI6P05 / AC9TDI8P06 — trace and run an algorithm someone else designed.
  • AC9M6P02 — use a system (the booklet) to run a simulation.
• Secondary
  • AC9TDI8K01 — explain how hardware specifications affect performance (here, the booklet’s “specifications” are corpus size and N value).
• General capabilities: Digital Literacy (Managing and operating).

Sampling#

Temperature and truncation strategies for controlling text output.

• Best fit: Years 7–10
• Primary
  • AC9M7P01 / AC9M8P02 — assigning probabilities to outcomes; combinations of events.
  • AC9TDI8P05 — design algorithms involving nested control structures (the truncation rules).
• Secondary
  • AC9M9P02 — relative frequencies from data to estimate probabilities.
• General capabilities: Critical and Creative Thinking (Analysing > Evaluating reasoning); Numeracy.

More Context#

Extend the bigram to two words of context (a trigram), see why that gets expensive, then add cheaper context with a skip grid.

• Best fit: Years 7–10
• Primary
  • AC9TDI8P03 — model and query the attributes of objects and events using structured data.
  • AC9TDI8P09 — implement, modify and debug programs involving control structures and functions.
• Secondary
  • AC9M8P02 — determine all possible combinations and use these to determine probabilities.
• General capabilities: Numeracy; Critical and Creative Thinking.

In-context Memory#

Add a short-term memory that biases generation toward recently-used words, keeping text on topic — a hands-on model of attention.

• Best fit: Years 7–10
• Primary
  • AC9TDI8P03 — model and query the attributes of objects and events using structured data (the running memory list).
  • AC9TDI8P09 — implement, modify and debug programs involving control structures and functions (the reweighting procedure over generation).
• Secondary
  • AC9M9P02 — use relative frequencies and adjust probabilities based on collected data (boosting recently-seen words).
• General capabilities: Critical and Creative Thinking (Analysing); Digital Literacy.

Induction Heads#

Complete a pattern from the text itself: find the last time the current word appeared and copy what followed — the circuit behind in-context learning.

• Best fit: Years 9–10
• Primary
  • AC9TDI8P06 / AC9TDI10P06 — trace algorithms to predict output, and validate by comparing output against test cases (the pattern-completion check).
  • AC9TDI8P09 — implement, modify and debug programs involving control structures and functions.
• Secondary
  • AC9TDI10P03 — model and query entities and their relationships using structured data.
• General capabilities: Critical and Creative Thinking (Inquiring, Analysing).

Word Embeddings#

Treat each word’s row in the bigram grid as a vector; measure similarity.

• Best fit: Years 9–10
• Primary
  • AC9M10SP02 — interpret networks and network diagrams used to represent relationships in practical situations. An embedding space is a relationship diagram.
  • AC9TDI10K02 — represent documents as content, structure and presentation; explain why such representations matter.
  • AC9TDI10P03 — model and query entities and their relationships using structured data.
• Secondary
  • AC9E9LA08 — vocabulary choices and meaning.
• General capabilities: Numeracy; Critical and Creative Thinking (Generating > Considering alternatives).

LoRA#

Add a lightweight adaptation layer to retarget a trained base model.

• Best fit: Years 9–10
• Primary
  • AC9TDI10P08 — generate, modify, communicate and critically evaluate alternative designs (the LoRA design choice itself).
  • AC9TDI10P06 — validate algorithms and programs by comparing output against a range of test cases.
• Secondary
  • AC9TDI10P09 — implement, modify and debug modular programs.
• General capabilities: Critical and Creative Thinking.

Synthetic Data#

Generate synthetic text, retrain on it, observe drift and collapse.

• Best fit: Years 9–10
• Primary
  • AC9TDI10P10 — evaluate solutions against design criteria, future impact, opportunities for enterprise. Model collapse is the case study.
  • AC9TDI10P01 — develop techniques to acquire, store and validate data (validation by retraining).
• Secondary
  • AC9TDI8P14 — assess whether collected data is essential to purpose.
• General capabilities: Critical and Creative Thinking (Reflecting); Ethical Understanding (Responsibility and accountability for action).

Agentic AI#

Turn the model into an agent by giving it access to external tools.

• Best fit: Years 9–10
• Primary
  • AC9TDI10P09 — implement modular programs applying algorithms and data structures (the tool-dispatch loop).
  • AC9TDI10P03 — model entities and their relationships (the model + tool + environment graph).
• Secondary
  • AC9TDI8P09 — programs involving control structures and functions.
• General capabilities: Digital Literacy (Investigating > Locate, generate and access data); Critical and Creative Thinking.

RLHF#

Use human preferences to adjust the model’s weights.

• Best fit: Years 9–10
• Primary
  • AC9TDI10P10 — evaluate solutions against design criteria, possible future impact and opportunities for enterprise.
  • AC9TDI8P10 — evaluate existing and student solutions against the design criteria and possible future impact.
• Secondary
  • AC9M9P02 — adjust probabilities based on collected data.
• General capabilities: Ethical Understanding (Decision-making and actions); Critical and Creative Thinking (Reflecting > Thinking about thinking).

Sycophancy#

Skew the model toward over-agreeable output by piling in flattering training data.

• Best fit: Years 9–10
• Primary
  • AC9TDI8P02 / AC9TDI10P02 — analyse and visualise data to draw conclusions and make predictions (here, predicting how a model will behave from its training set).
  • AC9TDI10P14 — apply the Australian Privacy Principles to critique digital systems (the lesson extends naturally into data-source critique).
• Secondary
  • AC9E10LY03 — analyse and evaluate how language features are used to implicitly or explicitly represent values, beliefs and attitudes.
• General capabilities: Ethical Understanding (Understanding ethical concepts > Examining values); Critical and Creative Thinking.

Coverage summary by strand#

Digital Technologies#

Most concentrated coverage. Across the lesson set, students engage substantively with these strands:

• Knowledge and understanding > Representation of data: covered by Training, More Context, In-context Memory, Word Embeddings.
• Processes and production skills > Acquiring, storing and validating data: Training, Synthetic Data, Sycophancy.
• Processes and production skills > Analysing and visualising data: Training, Sycophancy.
• Processes and production skills > Defining problems: Pre-trained Generation, Sampling.
• Processes and production skills > Designing and tracing algorithms: Generation, Sampling.
• Processes and production skills > Implementing programs: Generation, More Context, In-context Memory, Induction Heads, Agentic AI, LoRA.
• Processes and production skills > Evaluating solutions: RLHF, Synthetic Data, Sycophancy, LoRA.

The set does not substantively cover: networks and protocols, cyber security threat models, user-interface design, or project management. Those content descriptions still need a separate teaching vehicle.

Mathematics#

• Probability (Years 5–10): Weighted Randomness, Generation, Sampling, More Context, In-context Memory, RLHF.
• Statistics (Years 5–8): Training (categorising and describing data).
• Number / Algebra: light touch only.
• Space: only via Word Embeddings (vectors and distance).

English#

• Language strand (vocabulary): a light touch only — the materials no longer include a dedicated sentence-structure vehicle; Training (specialist vocabulary) and Word Embeddings (vocabulary and meaning) contribute.
• Literature / Literacy strands: light touch, mostly through critical analysis of generated vs. source text in Generation, Sycophancy, Synthetic Data.

General capabilities#

LLMs Unplugged is unusually rich in GC coverage because the activities require students to do the thinking, not just learn about it.

• Digital Literacy — the entire course is an extended worked example of “managing and operating” digital systems concepts without screens, with particular strength in Investigating > Locate, generate and access data and Managing and operating > Manage content, data and information.
• Critical and Creative Thinking — Inquiring, Generating ideas, possibilities and actions, Analysing, synthesising and evaluating reasoning and procedures, and Reflecting are all exercised across the lesson sequence.
• Numeracy — Recognising and using patterns and relationships, Interpreting and representing data, and Using probabilistic thinking underpin the fundamentals lessons.
• Literacy — Composing texts (generated output) and Comprehending texts (source corpus selection and analysis) are present throughout.
• Ethical Understanding — RLHF, Sycophancy, and Synthetic Data are designed to surface ethical issues around training data, feedback, and data quality.

Suggested programs#

Three concrete sequences that an administrator could slot into a year-long program.

Year 5–6: “How computers make text” (4 lessons, ~5 hours)#

1. Weighted Randomness — AC9M6P02
2. Training — AC9TDI6P05
3. Generation — AC9M6P02, AC9TDI6P05
4. Pre-trained Model Generation — AC9TDI6P05

Sits cleanly in a Digital Technologies unit on data representation and algorithms; Mathematics teacher can co-teach the probability components.

Year 7–8 Digital Technologies: “Data and AI” (6 lessons, ~8 hours)#

1. Weighted Randomness
2. Training — AC9TDI8P01, AC9TDI8P02
3. Generation — AC9TDI8P06
4. Sampling — AC9TDI8P05
5. More Context — AC9TDI8P03, AC9TDI8P09
6. In-context Memory — AC9TDI8P03, AC9TDI8P09

Delivers most of the “Data and information” knowledge band plus the algorithm-design content descriptions for Years 7–8.

Year 9–10 cross-curricular: “Understanding AI” (5 lessons, ~7 hours)#

1. More Context (recap, if not previously seen)
2. Word Embeddings — AC9TDI10K02, AC9TDI10P03
3. Synthetic Data — AC9TDI10P10, AC9TDI10P01
4. RLHF — AC9TDI10P10 + Ethical Understanding
5. Sycophancy — AC9TDI10P14 + Ethical Understanding

Co-teach with English (analysing language and values) and HASS / Civics (data ethics and decision-making).

Competency reference#

Every code cited above, grouped by learning area and strand. Codes follow the ACARA v9 pattern: AC9 + learning area (M Mathematics, TDI Digital Technologies, E English) + year level + strand (P probability, ST statistics, SP space, K knowledge, LA language, LY literacy) + sequence number.

Mathematics#

Probability:

• AC9M5P02 (Year 5) — conduct repeated chance experiments with equally and not equally likely outcomes; use frequency to compare and estimate likelihood.
• AC9M6P01 (Year 6) — probabilities lie on numerical scales; use fractions, decimals and percentages.
• AC9M6P02 (Year 6) — run simulations with increasing trials; compare observed with expected results.
• AC9M7P01 (Year 7) — identify sample spaces and assign probabilities to outcomes.
• AC9M7P02 (Year 7) — conduct repeated chance experiments and run simulations; compare predictions with observed results.
• AC9M8P02 (Year 8) — determine all possible combinations and use these to determine probabilities.
• AC9M9P02 (Year 9) — use relative frequencies from data to estimate probabilities; adjust probabilities based on collected data.

Statistics:

• AC9M5ST01 (Year 5) — categorise and describe data and frequencies.
• AC9M6ST01 (Year 6) — categorise and describe data and frequencies.

Space:

• AC9M10SP02 (Year 10) — interpret networks and network diagrams used to represent relationships in practical situations.

Digital Technologies#

Knowledge and understanding:

• AC9TDI6K03 (Year 6) — digital systems represent data using numbers.
• AC9TDI8K01 (Year 8) — explain how hardware specifications affect performance.
• AC9TDI10K02 (Year 10) — represent documents as content, structure and presentation; explain why such representations matter.

Processes and production skills:

• AC9TDI6P05 (Year 6) — implement algorithms involving control structures, variables and input.
• AC9TDI8P01 (Year 8) — acquire, store and validate data from a range of sources.
• AC9TDI8P02 (Year 8) — analyse and visualise data to identify patterns.
• AC9TDI8P03 (Year 8) — model and query the attributes of objects and events using structured data.
• AC9TDI8P05 (Year 8) — design algorithms involving nested control structures.
• AC9TDI8P06 (Year 8) — trace algorithms to predict output for a given input.
• AC9TDI8P09 (Year 8) — implement, modify and debug programs involving control structures and functions.
• AC9TDI8P10 (Year 8) — evaluate existing and student solutions against the design criteria and possible future impact.
• AC9TDI8P14 (Year 8) — assess whether collected data is essential to purpose.
• AC9TDI10P01 (Year 10) — develop techniques to acquire, store and validate data.
• AC9TDI10P02 (Year 10) — analyse and visualise data to draw conclusions and make predictions.
• AC9TDI10P03 (Year 10) — model and query entities and their relationships using structured data.
• AC9TDI10P06 (Year 10) — validate algorithms and programs by comparing output against a range of test cases.
• AC9TDI10P08 (Year 10) — generate, modify, communicate and critically evaluate alternative designs.
• AC9TDI10P09 (Year 10) — implement, modify and debug modular programs applying algorithms and data structures.
• AC9TDI10P10 (Year 10) — evaluate solutions against design criteria, possible future impact and opportunities for enterprise.
• AC9TDI10P14 (Year 10) — apply the Australian Privacy Principles to critique digital systems.

English#

Language:

• AC9E5LA03 (Year 5) — describe how texts use language features and are organised into stages and phases.
• AC9E7LA08 (Year 7) — vocabulary in building specialist knowledge.
• AC9E9LA08 (Year 9) — analyse how vocabulary choices contribute to style, mood and tone.

Literacy:

• AC9E10LY03 (Year 10) — analyse and evaluate how language features are used to implicitly or explicitly represent values, beliefs and attitudes.

Caveats and out of scope#

• Foundation to Year 4: not currently mapped. The dice-and-grid mechanics are accessible to upper primary, but the content descriptions at F–4 emphasise concrete materials, not symbolic systems; we recommend treating any F–4 use as enrichment rather than as primary curriculum coverage.
• Senior secondary (11–12): out of scope. Some lessons (especially Synthetic Data, RLHF, Word Embeddings) would extend naturally into the Digital Solutions / Software Engineering / Specialist Mathematics syllabi, but those syllabi are state-specific and not mapped here.
• Assessment: this document maps teaching coverage, not assessment evidence. Each primary mapping above could plausibly serve as evidence, but the assessment task design (rubrics, work samples, moderation) is a separate exercise.
• Codes verified against ACARA workbook of 2026. If ACARA publishes a v9.x update, codes should be re-checked; the structural mapping (which lesson aligns with which strand) is unlikely to change.

Next steps#

1. External review by a practising secondary DT teacher and a Mathematics teacher.
2. Sample lesson plans for each of the three suggested programs above, including assessment artefacts.
3. State-level cross-references (NESA Stage 4/5, VCAA Levels 7–10) once the F–10 mapping is settled.