CRADLE Bench: A Clinician-Annotated Benchmark for Multi-Faceted Mental Health Crisis and Safety Risk Detection
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909998545633280 |
|---|---|
| author | Byun, Grace Lipschutz, Rebecca Minton, Sean T. Lott, Abigail Choi, Jinho D. |
| author_facet | Byun, Grace Lipschutz, Rebecca Minton, Sean T. Lott, Abigail Choi, Jinho D. |
| contents | Detecting mental health crisis situations such as suicide ideation, rape, domestic violence, child abuse, and sexual harassment is a critical yet underexplored challenge for language models. When such situations arise during user--model interactions, models must reliably flag them, as failure to do so can have serious consequences. In this work, we introduce CRADLE BENCH, a benchmark for multi-faceted crisis detection. Unlike previous efforts that focus on a limited set of crisis types, our benchmark covers seven types defined in line with clinical standards and is the first to incorporate temporal labels. Our benchmark provides 600 clinician-annotated evaluation examples and 420 development examples, together with a training corpus of around 4K examples automatically labeled using a majority-vote ensemble of multiple language models, which significantly outperforms single-model annotation. We further fine-tune six crisis detection models on subsets defined by consensus and unanimous ensemble agreement, providing complementary models trained under different agreement criteria. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_23845 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | CRADLE Bench: A Clinician-Annotated Benchmark for Multi-Faceted Mental Health Crisis and Safety Risk Detection Byun, Grace Lipschutz, Rebecca Minton, Sean T. Lott, Abigail Choi, Jinho D. Computation and Language Artificial Intelligence Detecting mental health crisis situations such as suicide ideation, rape, domestic violence, child abuse, and sexual harassment is a critical yet underexplored challenge for language models. When such situations arise during user--model interactions, models must reliably flag them, as failure to do so can have serious consequences. In this work, we introduce CRADLE BENCH, a benchmark for multi-faceted crisis detection. Unlike previous efforts that focus on a limited set of crisis types, our benchmark covers seven types defined in line with clinical standards and is the first to incorporate temporal labels. Our benchmark provides 600 clinician-annotated evaluation examples and 420 development examples, together with a training corpus of around 4K examples automatically labeled using a majority-vote ensemble of multiple language models, which significantly outperforms single-model annotation. We further fine-tune six crisis detection models on subsets defined by consensus and unanimous ensemble agreement, providing complementary models trained under different agreement criteria. |
| title | CRADLE Bench: A Clinician-Annotated Benchmark for Multi-Faceted Mental Health Crisis and Safety Risk Detection |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2510.23845 |