WorkMotionCase file 02, 3 of 3 in this programme
An assessment system where code calculates every number
A movement-assessment system where Python calculates every number and the language model only writes the explanation. A second pipeline swaps those roles.
- Where
- IdeasLab Formosa, 2025 — present
- My part
- Architect: rule engine, scoring, prompts, evaluation
- Built with
- LLM tool use, RAG, Rule engines, Python, Structured prompting
An incorrect measurement in an assessment can change what someone does next, so the system must prevent the language model from inventing numbers.
- 0
- numbers calculated by the language model
- 8
- languages, prompts included
4 more sections
The rest of this study is private
Reach out on LinkedIn or by email for research collaboration or a deeper technical discussion.
The core rule
Same input, two pipelines
Testing the rulebook itself
Grounding and cost
What it taught
Let code do the arithmetic
All the lessonsWhen a wrong number can hurt someone, Python calculates it and the language model only explains it.