Risk Aware Benchmarking of Large Language Models

• URL: http://arxiv.org/abs/2310.07132v3
• Date: Sun, 9 Jun 2024 18:26:34 GMT
• Title: Risk Aware Benchmarking of Large Language Models
• Authors: Apoorva Nitsure, Youssef Mroueh, Mattia Rigotti, Kristjan Greenewald, Brian Belgodere, Mikhail Yurochkin, Jiri Navratil, Igor Melnyk, Jerret Ross,
• Abstract summary: We propose a distributional framework for benchmarking socio-technical risks of foundation models with quantified statistical significance. We show that the second order statistics in this test are linked to mean-risk models commonly used in econometrics and mathematical finance. We use our framework to compare various large language models regarding risks related to drifting from instructions and outputting toxic content.
• Score: 36.95053112313244
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: 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.

Related papers

• Multivariate Stochastic Dominance via Optimal Transport and Applications to Models Benchmarking [21.23500484100963]
  We introduce a statistic that assesses almost dominance under the framework of Optimal Transport with a smooth cost. We also propose a hypothesis testing framework as well as an efficient implementation using the Sinkhorn algorithm. We showcase our method in comparing and benchmarking Large Language Models that are evaluated on multiple metrics.
  arXiv  Detail & Related papers  (2024-06-10T16:14:50Z)
• Likelihood Ratio Confidence Sets for Sequential Decision Making [51.66638486226482]
  We revisit the likelihood-based inference principle and propose to use likelihood ratios to construct valid confidence sequences. Our method is especially suitable for problems with well-specified likelihoods. We show how to provably choose the best sequence of estimators and shed light on connections to online convex optimization.
  arXiv  Detail & Related papers  (2023-11-08T00:10:21Z)
• Distribution-free Deviation Bounds and The Role of Domain Knowledge in Learning via Model Selection with Cross-validation Risk Estimation [0.0]
  Cross-validation techniques for risk estimation and model selection are widely used in statistics and machine learning. This paper presents learning via model selection with cross-validation risk estimation as a general systematic learning framework.
  arXiv  Detail & Related papers  (2023-03-15T17:18:31Z)
• On (assessing) the fairness of risk score models [2.0646127669654826]
  Risk models are of interest for a number of reasons, including the fact that they communicate uncertainty about the potential outcomes to users. We identify the provision of similar value to different groups as a key desideratum for risk score fairness. We introduce a novel calibration error metric that is less sample size-biased than previously proposed metrics.
  arXiv  Detail & Related papers  (2023-02-17T12:45:51Z)
• MAUVE Scores for Generative Models: Theory and Practice [95.86006777961182]
  We present MAUVE, a family of comparison measures between pairs of distributions such as those encountered in the generative modeling of text or images. We find that MAUVE can quantify the gaps between the distributions of human-written text and those of modern neural language models. We demonstrate in the vision domain that MAUVE can identify known properties of generated images on par with or better than existing metrics.
  arXiv  Detail & Related papers  (2022-12-30T07:37:40Z)
• Causality and Generalizability: Identifiability and Learning Methods [0.0]
  This thesis contributes to the research areas concerning the estimation of causal effects, causal structure learning, and distributionally robust prediction methods. We present novel and consistent linear and non-linear causal effects estimators in instrumental variable settings that employ data-dependent mean squared prediction error regularization. We propose a general framework for distributional robustness with respect to intervention-induced distributions.
  arXiv  Detail & Related papers  (2021-10-04T13:12:11Z)
• Adaptive learning for financial markets mixing model-based and model-free RL for volatility targeting [0.0]
  Model-Free Reinforcement Learning has achieved meaningful results in stable environments but, to this day, it remains problematic in regime changing environments like financial markets. We propose to combine the best of the two techniques by selecting various model-based approaches thanks to Model-Free Deep Reinforcement Learning.
  arXiv  Detail & Related papers  (2021-04-19T19:20:22Z)
• Characterizing Fairness Over the Set of Good Models Under Selective Labels [69.64662540443162]
  We develop a framework for characterizing predictive fairness properties over the set of models that deliver similar overall performance. We provide tractable algorithms to compute the range of attainable group-level predictive disparities. We extend our framework to address the empirically relevant challenge of selectively labelled data.
  arXiv  Detail & Related papers  (2021-01-02T02:11:37Z)
• Trust but Verify: Assigning Prediction Credibility by Counterfactual Constrained Learning [123.3472310767721]
  Prediction credibility measures are fundamental in statistics and machine learning. These measures should account for the wide variety of models used in practice. The framework developed in this work expresses the credibility as a risk-fit trade-off.
  arXiv  Detail & Related papers  (2020-11-24T19:52:38Z)
• Decision-Making with Auto-Encoding Variational Bayes [71.44735417472043]
  We show that a posterior approximation distinct from the variational distribution should be used for making decisions. Motivated by these theoretical results, we propose learning several approximate proposals for the best model. In addition to toy examples, we present a full-fledged case study of single-cell RNA sequencing.
  arXiv  Detail & Related papers  (2020-02-17T19:23:36Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.