Risk Aware Benchmarking of Large Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Nitsure, Apoorva, Mroueh, Youssef, Rigotti, Mattia, Greenewald, Kristjan, Belgodere, Brian, Yurochkin, Mikhail, Navratil, Jiri, Melnyk, Igor, Ross, Jerret
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909220013604864
author Nitsure, Apoorva
Mroueh, Youssef
Rigotti, Mattia
Greenewald, Kristjan
Belgodere, Brian
Yurochkin, Mikhail
Navratil, Jiri
Melnyk, Igor
Ross, Jerret
author_facet Nitsure, Apoorva
Mroueh, Youssef
Rigotti, Mattia
Greenewald, Kristjan
Belgodere, Brian
Yurochkin, Mikhail
Navratil, Jiri
Melnyk, Igor
Ross, Jerret
contents We propose a distributional framework for benchmarking socio-technical risks of foundation models with quantified statistical significance. Our approach hinges on a new statistical relative testing based on first and second order stochastic dominance of real random variables. We show that the second order statistics in this test are linked to mean-risk models commonly used in econometrics and mathematical finance to balance risk and utility when choosing between alternatives. Using this framework, we formally develop a risk-aware approach for foundation model selection given guardrails quantified by specified metrics. Inspired by portfolio optimization and selection theory in mathematical finance, we define a metrics portfolio for each model as a means to aggregate a collection of metrics, and perform model selection based on the stochastic dominance of these portfolios. The statistical significance of our tests is backed theoretically by an asymptotic analysis via central limit theorems instantiated in practice via a bootstrap variance estimate. We use our framework to compare various large language models regarding risks related to drifting from instructions and outputting toxic content.
format Preprint
id arxiv_https___arxiv_org_abs_2310_07132
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Risk Aware Benchmarking of Large Language Models
Nitsure, Apoorva
Mroueh, Youssef
Rigotti, Mattia
Greenewald, Kristjan
Belgodere, Brian
Yurochkin, Mikhail
Navratil, Jiri
Melnyk, Igor
Ross, Jerret
Machine Learning
Statistics Theory
Risk Management
We propose a distributional framework for benchmarking socio-technical risks of foundation models with quantified statistical significance. Our approach hinges on a new statistical relative testing based on first and second order stochastic dominance of real random variables. We show that the second order statistics in this test are linked to mean-risk models commonly used in econometrics and mathematical finance to balance risk and utility when choosing between alternatives. Using this framework, we formally develop a risk-aware approach for foundation model selection given guardrails quantified by specified metrics. Inspired by portfolio optimization and selection theory in mathematical finance, we define a metrics portfolio for each model as a means to aggregate a collection of metrics, and perform model selection based on the stochastic dominance of these portfolios. The statistical significance of our tests is backed theoretically by an asymptotic analysis via central limit theorems instantiated in practice via a bootstrap variance estimate. We use our framework to compare various large language models regarding risks related to drifting from instructions and outputting toxic content.
title Risk Aware Benchmarking of Large Language Models
topic Machine Learning
Statistics Theory
Risk Management
url https://arxiv.org/abs/2310.07132