# BAGEL preprint reports prompt defenses built from small classifiers

_Published Wednesday, October 21, 2026 at 3:06 AM EDT · Science, AI · Latest · Tier 2 — Notable_

Researchers report that BAGEL, an ensemble of small classifiers that detects malicious prompts, outperformed popular moderation APIs and most tested white-box defenses in a single-turn evaluation. The preprint reports an F1 score of 0.757 with an ensemble equivalent to 430M parameters, or 0.772 with a learned router and 324M parameters.

BAGEL combines predictions from selected classifiers trained on different attack datasets. New attacks can be addressed by training and adding one 86M-parameter classifier rather than retraining the whole ensemble. Evaluation used held-out datasets excluded from training and calibration. Its multi-turn F1 score was 0.660, below its single-turn results but above the billion-parameter methods tested.

## Sources

- [cs.LG updates on arXiv.org](https://arxiv.org/abs/2602.08062)

---
Canonical: https://techandbusiness.org/newswire/kqIcGwmR6nyHHQl8Cm1kTh
Published: 2026-10-21T07:06:45.143Z
Story chronology: 2026-10-21T04:00:00.000Z
Retrieved: 2026-10-25T06:21:46.414Z
Publisher: Tech & Business (techandbusiness.org)