A Probabilistic Perspective on Unlearning and Alignment for Large Language Models

• URL: http://arxiv.org/abs/2410.03523v5
• Date: Sat, 15 Feb 2025 21:53:44 GMT
• Title: A Probabilistic Perspective on Unlearning and Alignment for Large Language Models
• Authors: Yan Scholten, Stephan Günnemann, Leo Schwinn,
• Abstract summary: We introduce the first formal probabilistic evaluation framework for Large Language Models (LLMs) Namely, we propose novel metrics with high probability guarantees concerning the output distribution of a model. Our metrics are application-independent and allow practitioners to make more reliable estimates about model capabilities before deployment.
• Score: 48.96686419141881
• License:
• Abstract: Comprehensive evaluation of Large Language Models (LLMs) is an open research problem. Existing evaluations rely on deterministic point estimates generated via greedy decoding. However, we find that deterministic evaluations fail to capture the whole output distribution of a model, yielding inaccurate estimations of model capabilities. This is particularly problematic in critical contexts such as unlearning and alignment, where precise model evaluations are crucial. To remedy this, we introduce the first formal probabilistic evaluation framework for LLMs. Namely, we propose novel metrics with high probability guarantees concerning the output distribution of a model. Our metrics are application-independent and allow practitioners to make more reliable estimates about model capabilities before deployment. Our experimental analysis reveals that deterministic evaluations falsely indicate successful unlearning and alignment, whereas our probabilistic evaluations better capture model capabilities. We show how to overcome challenges associated with probabilistic outputs in a case study on unlearning by introducing (1) a novel loss based on entropy optimization, and (2) adaptive temperature scaling. We demonstrate that our approach significantly enhances unlearning in probabilistic settings on recent benchmarks. Overall, our proposed shift from point estimates to probabilistic evaluations of output distributions represents an important step toward comprehensive evaluations of LLMs. Code available at https://www.cs.cit.tum.de/daml/probabilistic-unlearning/.

Related papers

• Model-free Methods for Event History Analysis and Efficient Adjustment (PhD Thesis) [55.2480439325792]
  This thesis is a series of independent contributions to statistics unified by a model-free perspective. The first chapter elaborates on how a model-free perspective can be used to formulate flexible methods that leverage prediction techniques from machine learning. The second chapter studies the concept of local independence, which describes whether the evolution of one process is directly influenced by another.
  arXiv  Detail & Related papers  (2025-02-11T19:24:09Z)
• Deep Probability Segmentation: Are segmentation models probability estimators? [0.7646713951724011]
  We apply Calibrated Probability Estimation to segmentation tasks to evaluate its impact on model calibration. Results indicate that while CaPE improves calibration, its effect is less pronounced compared to classification tasks. We also investigated the influence of dataset size and bin optimization on the effectiveness of calibration.
  arXiv  Detail & Related papers  (2024-09-19T07:52:19Z)
• Rethinking Evaluation Metric for Probability Estimation Models Using Esports Data [8.10304644344495]
  We propose a novel metric called Balance score which is a simple yet effective metric in terms of six good properties that probability estimation metric should have. Under the general condition, we also found that the Balance score can be an effective approximation of the true expected calibration error.
  arXiv  Detail & Related papers  (2023-09-12T14:04:12Z)
• Value-Distributional Model-Based Reinforcement Learning [59.758009422067]
  Quantifying uncertainty about a policy's long-term performance is important to solve sequential decision-making tasks. We study the problem from a model-based Bayesian reinforcement learning perspective. We propose Epistemic Quantile-Regression (EQR), a model-based algorithm that learns a value distribution function.
  arXiv  Detail & Related papers  (2023-08-12T14:59:19Z)
• Plan To Predict: Learning an Uncertainty-Foreseeing Model for Model-Based Reinforcement Learning [32.24146877835396]
  We propose emphPlan To Predict (P2P), a framework that treats the model rollout process as a sequential decision making problem. We show that P2P achieves state-of-the-art performance on several challenging benchmark tasks.
  arXiv  Detail & Related papers  (2023-01-20T10:17:22Z)
• The Implicit Delta Method [61.36121543728134]
  In this paper, we propose an alternative, the implicit delta method, which works by infinitesimally regularizing the training loss of uncertainty. We show that the change in the evaluation due to regularization is consistent for the variance of the evaluation estimator, even when the infinitesimal change is approximated by a finite difference.
  arXiv  Detail & Related papers  (2022-11-11T19:34:17Z)
• BRIO: Bringing Order to Abstractive Summarization [107.97378285293507]
  We propose a novel training paradigm which assumes a non-deterministic distribution. Our method achieves a new state-of-the-art result on the CNN/DailyMail (47.78 ROUGE-1) and XSum (49.07 ROUGE-1) datasets.
  arXiv  Detail & Related papers  (2022-03-31T05:19:38Z)
• Variance based sensitivity analysis for Monte Carlo and importance sampling reliability assessment with Gaussian processes [0.0]
  We propose a methodology to quantify the sensitivity of the probability of failure estimator to two uncertainty sources. This analysis also enables to control the whole error associated to the failure probability estimate and thus provides an accuracy criterion on the estimation. The approach is proposed for both a Monte Carlo based method as well as an importance sampling based method, seeking to improve the estimation of rare event probabilities.
  arXiv  Detail & Related papers  (2020-11-30T17:06:28Z)
• Bootstrapped model learning and error correction for planning with uncertainty in model-based RL [1.370633147306388]
  A natural aim is to learn a model that reflects accurately the dynamics of the environment. This paper explores the problem of model misspecification through uncertainty-aware reinforcement learning agents. We propose a bootstrapped multi-headed neural network that learns the distribution of future states and rewards.
  arXiv  Detail & Related papers  (2020-04-15T15:41:21Z)
• Meta-Learned Confidence for Few-shot Learning [60.6086305523402]
  A popular transductive inference technique for few-shot metric-based approaches, is to update the prototype of each class with the mean of the most confident query examples. We propose to meta-learn the confidence for each query sample, to assign optimal weights to unlabeled queries. We validate our few-shot learning model with meta-learned confidence on four benchmark datasets.
  arXiv  Detail & Related papers  (2020-02-27T10:22:17Z)

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.