Latent Ability
A browsable workshop site for the latent-ability-aware evaluation material, organized by section, activities, and supporting notes.
The workshop now progresses through three connected blocks. The opening block joins supervised evaluation with Binary IRT and 2PL: first we see that examples are not equally difficult, then we introduce a latent language for ability, difficulty, and discrimination. From there, the workshop moves to Beta4-IRT, and the final block begins with unsupervised evaluation before folding that intuition directly into CLAIRE.
Workshop Structure
Section 1
Supervised evaluation as motivation, followed by Binary IRT, 1PL intuition, 2PL-IRT, and ICC interpretation.
Section 2
Beta4-IRT for bounded responses, synthetic recovery, and latent quantities beyond the binary case.
Section 3
CLAIRE, starting from unsupervised disagreement and ending with agreement-based latent evaluation of clustering models.
Suggested Rhythm
A realistic way to run this workshop in a 3-hour slot is to group the material into three classroom blocks:
Section 1: supervised evaluation as motivation, then Binary IRT and 2PL.Section 2: Beta4-IRT as the bounded-response extension.Section 3: a CLAIRE block that opens with unsupervised evaluation and then turns agreement into the latent framework.
For each block, keep the same classroom cadence:
15 minutesof theory15 minutesof guided hands-on explanation30 minutesof participant activity
That gives:
1 hourfor supervised evaluation plus Binary IRT and 2PL1 hourfor Beta4-IRT1 hourfor the integrated CLAIRE block
Total estimated duration: 3 hours.
Repository
- Main repository: latent-ability-ml
- Roadmap notebook: 00_workshop_roadmap.ipynb
- Portuguese script: 99_script_portuguese.md
- English script: 99_script_english.md
Guide
The workshop guide is also available through the site navigation, so participants can access the repository overview without leaving the website.
