prima·bench
Repo ↗

Cell-level evals · eval bucket

Identity, state, response.

Identity, state, and response to perturbation — reading a cell from its measured profile.

24 evals implemented · 3 families
01

Overview

The cell bucket grades a model on a cell's identity, state, and how it responds to perturbation. 24 evals span 3 task families, each a (task × dataset-arm) pair drawn straight from the runtime registry.

02

Families

Task families present in this bucket, by eval count.

03

Catalog

Filter and sort every cell eval; expand any card for its IO contract.

24 of 24 evals · 3 families

integration15

  • 11m
  • 11m
  • 11m
  • 11m
  • 11m
  • 11m
  • 11m
  • 11m
  • 11m
  • 11m
  • 11m
  • 11m
  • 11m
  • 13m
  • 13m

cell_type_annotation8

  • 6m
  • 6m
  • 6m
  • 6m
  • 6m
  • 6m
  • 6m
  • 6m

perturbation1

  • 2m