01 · Visual explanation
Making ordinal models plot-able
Visual tools that connect model parameters to latent evidence, response thresholds, and uncertainty.
ML · Model evaluation · Research tools
Tomer Zipori
Playful questions. Rigorous answers.
I build models, experiments, and tools to understand how people—and machines—make sense of the world.
Illustrative model
Same evidence. Different decision.
Move a criterion. The response probabilities change; fixed discriminability does not.
Noise
Signal
d′ = 1.55 · criteria shift responses, not discriminability
Selected work
Questions with somewhere to go.
Three finished projects where a claim becomes a tool, a workflow, or a picture.
02 · Research workflow
Harness the power of your laptop
A toolkit for repository exploration, review, and candidate patches, with human verification built into the workflow.
03 · Reproducible analysis
Measuring a cinematic hunch
A Python pipeline for testing a claim about movie colour, with the movie as the unit of analysis.
Currently investigating
Can padding change what a model learns?
I’m testing whether padding-sensitive normalization changes audio-classification metrics in a small DistilHuBERT model. The study compares ordinary and masked normalization under two padding conditions.
Read the experiment note Ongoing study · Pilot complete
Garden
Follow a question.
These paths connect my projects, notes, and longer articles. Pick a theme, or take an unexpected turn.
Recent notes
Small distinctions with large consequences.
Notes are short, but they keep the measurement attached to the question.
Can padding change audio-classification metrics?
An ongoing DistilHuBERT study comparing ordinary and masked positional BatchNorm under minimal and additional padding.
Ongoing studyNote · 2026
A useful local agent earns its keep
Evaluate a local agent by the useful result after preparation, review, correction, and unsuccessful attempts.
Note · 2026
Criteria move the answers. Separation moves discriminability.
A worked example showing how decision thresholds change responses without changing fixed discriminability.
Note · 2026
About and contact
Let’s talk about models, experiments, and useful tools.
My background spans psychology, experimental methods, statistics, data science, and machine learning. I turn research questions into models, visual explanations, and tools that other people can inspect.
