WalledEval: A Comprehensive Safety Evaluation Toolkit for Large Language Models

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Gupta, Prannaya, Yau, Le Qi, Low, Hao Han, Lee, I-Shiang, Lim, Hugo Maximus, Teoh, Yu Xin, Koh, Jia Hng, Liew, Dar Win, Bhardwaj, Rishabh, Bhardwaj, Rajat, Poria, Soujanya
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910570882531328
author Gupta, Prannaya
Yau, Le Qi
Low, Hao Han
Lee, I-Shiang
Lim, Hugo Maximus
Teoh, Yu Xin
Koh, Jia Hng
Liew, Dar Win
Bhardwaj, Rishabh
Bhardwaj, Rajat
Poria, Soujanya
author_facet Gupta, Prannaya
Yau, Le Qi
Low, Hao Han
Lee, I-Shiang
Lim, Hugo Maximus
Teoh, Yu Xin
Koh, Jia Hng
Liew, Dar Win
Bhardwaj, Rishabh
Bhardwaj, Rajat
Poria, Soujanya
contents WalledEval is a comprehensive AI safety testing toolkit designed to evaluate large language models (LLMs). It accommodates a diverse range of models, including both open-weight and API-based ones, and features over 35 safety benchmarks covering areas such as multilingual safety, exaggerated safety, and prompt injections. The framework supports both LLM and judge benchmarking and incorporates custom mutators to test safety against various text-style mutations, such as future tense and paraphrasing. Additionally, WalledEval introduces WalledGuard, a new, small, and performant content moderation tool, and two datasets: SGXSTest and HIXSTest, which serve as benchmarks for assessing the exaggerated safety of LLMs and judges in cultural contexts. We make WalledEval publicly available at https://github.com/walledai/walledeval.
format Preprint
id arxiv_https___arxiv_org_abs_2408_03837
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle WalledEval: A Comprehensive Safety Evaluation Toolkit for Large Language Models
Gupta, Prannaya
Yau, Le Qi
Low, Hao Han
Lee, I-Shiang
Lim, Hugo Maximus
Teoh, Yu Xin
Koh, Jia Hng
Liew, Dar Win
Bhardwaj, Rishabh
Bhardwaj, Rajat
Poria, Soujanya
Computation and Language
Artificial Intelligence
WalledEval is a comprehensive AI safety testing toolkit designed to evaluate large language models (LLMs). It accommodates a diverse range of models, including both open-weight and API-based ones, and features over 35 safety benchmarks covering areas such as multilingual safety, exaggerated safety, and prompt injections. The framework supports both LLM and judge benchmarking and incorporates custom mutators to test safety against various text-style mutations, such as future tense and paraphrasing. Additionally, WalledEval introduces WalledGuard, a new, small, and performant content moderation tool, and two datasets: SGXSTest and HIXSTest, which serve as benchmarks for assessing the exaggerated safety of LLMs and judges in cultural contexts. We make WalledEval publicly available at https://github.com/walledai/walledeval.
title WalledEval: A Comprehensive Safety Evaluation Toolkit for Large Language Models
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2408.03837