LetTokenBurn
Sign in
Auto71%Harbor: pendingMultilingualKnowledge & Reasoning

CLARA

CLARA: Can AI Assess Developmental Appropriateness in Children's Stories?

A bilingual benchmark of 1107 Chinese-English children's stories with structured developmental annotations across cognitive, language, and social-emotional dimensions to evaluate AI alignment with human developmental judgments.

Paper

How to run

ltb · clara
$ npm i -g lettokenburn-cli
$ ltb bench info clara

External harness. No Harbor adapter has been generated for this benchmark yet. `ltb bench info` prints the known links.

Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs

Scores

4 sourced entries
HarnessSourceConfidenceLink
Gemini 2.5 Pro0.9N/AIndependentmediumOct 5, 2026arxiv.org
Qwen Max (2025-01-25)0.7N/AIndependentmediumOct 5, 2026arxiv.org
GPT-4o0.7N/AIndependentmediumOct 5, 2026arxiv.org
DeepSeek Chat0.7N/AIndependentmediumOct 5, 2026arxiv.org

Every number links to where it was published. Low-confidence rows are muted and need per-cell verification.

Reproduced by the community

All runs
Loading community runs…

About

The benchmark contains 1107 aligned Chinese-English children's stories with normalized silver developmental references and structured annotations. It evaluates AI systems' ability to approximate human developmental judgments through structured annotation rather than surface readability.

Scoring involves comparing model outputs against developmental references and human educator rankings. The best reported overlap accuracy is 0.904, with a mean developmental distance of 0.096 and 0.81 agreement with human rankings.