JurEE not Judges: safeguarding llm interactions with small, specialised Encoder Ensembles

TL;DR

JurEE is an ensemble of specialized encoder-only transformer models that provides probabilistic risk assessments for AI-User interactions, outperforming existing methods in accuracy, speed, and cost-efficiency for content moderation tasks.

Contribution

This paper introduces JurEE, a novel ensemble of encoder-only transformers that offers robust, interpretable, and efficient risk estimation across diverse safety scenarios in LLM-based systems.

Findings

JurEE significantly outperforms baseline models in accuracy and speed.

JurEE demonstrates superior cost-efficiency for large-scale moderation.

The modular design allows customizable risk thresholds for various applications.

Abstract

We introduce JurEE, an ensemble of efficient, encoder-only transformer models designed to strengthen safeguards in AI-User interactions within LLM-based systems. Unlike existing LLM-as-Judge methods, which often struggle with generalization across risk taxonomies and only provide textual outputs, JurEE offers probabilistic risk estimates across a wide range of prevalent risks. Our approach leverages diverse data sources and employs progressive synthetic data generation techniques, including LLM-assisted augmentation, to enhance model robustness and performance. We create an in-house benchmark comprising of other reputable benchmarks such as the OpenAI Moderation Dataset and ToxicChat, where we find JurEE significantly outperforms baseline models, demonstrating superior accuracy, speed, and cost-efficiency. This makes it particularly suitable for applications requiring stringent content moderation, such as customer-facing chatbots. The encoder-ensemble's modular design allows users to set tailored risk thresholds, enhancing its versatility across various safety-related applications. JurEE's collective decision-making process, where each specialized encoder model contributes to the final output, not only improves predictive accuracy but also enhances interpretability. This approach provides a more efficient, performant, and economical alternative to traditional LLMs for large-scale implementations requiring robust content moderation.