Language Model Cascades: Token-level uncertainty and beyond
- URL: http://arxiv.org/abs/2404.10136v1
- Date: Mon, 15 Apr 2024 21:02:48 GMT
- Title: Language Model Cascades: Token-level uncertainty and beyond
- Authors: Neha Gupta, Harikrishna Narasimhan, Wittawat Jitkrittum, Ankit Singh Rawat, Aditya Krishna Menon, Sanjiv Kumar,
- Abstract summary: Recent advances in language models (LMs) have led to significant improvements in quality on complex NLP tasks.
Cascading offers a simple strategy to achieve more favorable cost-quality tradeoffs.
We show that incorporating token-level uncertainty through learned post-hoc deferral rules can significantly outperform simple aggregation strategies.
- Score: 65.38515344964647
- License: http://creativecommons.org/licenses/by/4.0/
- Abstract: Recent advances in language models (LMs) have led to significant improvements in quality on complex NLP tasks, but at the expense of increased inference costs. Cascading offers a simple strategy to achieve more favorable cost-quality tradeoffs: here, a small model is invoked for most "easy" instances, while a few "hard" instances are deferred to the large model. While the principles underpinning cascading are well-studied for classification tasks - with deferral based on predicted class uncertainty favored theoretically and practically - a similar understanding is lacking for generative LM tasks. In this work, we initiate a systematic study of deferral rules for LM cascades. We begin by examining the natural extension of predicted class uncertainty to generative LM tasks, namely, the predicted sequence uncertainty. We show that this measure suffers from the length bias problem, either over- or under-emphasizing outputs based on their lengths. This is because LMs produce a sequence of uncertainty values, one for each output token; and moreover, the number of output tokens is variable across examples. To mitigate this issue, we propose to exploit the richer token-level uncertainty information implicit in generative LMs. We argue that naive predicted sequence uncertainty corresponds to a simple aggregation of these uncertainties. By contrast, we show that incorporating token-level uncertainty through learned post-hoc deferral rules can significantly outperform such simple aggregation strategies, via experiments on a range of natural language benchmarks with FLAN-T5 models. We further show that incorporating embeddings from the smaller model and intermediate layers of the larger model can give an additional boost in the overall cost-quality tradeoff.
Related papers
- Quantifying Prediction Consistency Under Model Multiplicity in Tabular LLMs [10.494477811252034]
Fine-tuning large language models can lead to textitfine-tuning multiplicity, where equally well-performing models make conflicting predictions on the same inputs.
This raises critical concerns about the robustness and reliability of Tabular LLMs.
This work proposes a novel metric to quantify the robustness of individual predictions without expensive model retraining.
arXiv Detail & Related papers (2024-07-04T22:22:09Z) - Deep Bayesian Active Learning for Preference Modeling in Large Language Models [84.817400962262]
We propose the Bayesian Active Learner for Preference Modeling (BAL-PM) for Preference Modeling.
BAL-PM requires 33% to 68% fewer preference labels in two popular human preference datasets and exceeds previous Bayesian acquisition policies.
Our experiments demonstrate that BAL-PM requires 33% to 68% fewer preference labels in two popular human preference datasets and exceeds previous Bayesian acquisition policies.
arXiv Detail & Related papers (2024-06-14T13:32:43Z) - Cycles of Thought: Measuring LLM Confidence through Stable Explanations [53.15438489398938]
Large language models (LLMs) can reach and even surpass human-level accuracy on a variety of benchmarks, but their overconfidence in incorrect responses is still a well-documented failure mode.
We propose a framework for measuring an LLM's uncertainty with respect to the distribution of generated explanations for an answer.
arXiv Detail & Related papers (2024-06-05T16:35:30Z) - Low-rank finetuning for LLMs: A fairness perspective [54.13240282850982]
Low-rank approximation techniques have become the de facto standard for fine-tuning Large Language Models.
This paper investigates the effectiveness of these methods in capturing the shift of fine-tuning datasets from the initial pre-trained data distribution.
We show that low-rank fine-tuning inadvertently preserves undesirable biases and toxic behaviors.
arXiv Detail & Related papers (2024-05-28T20:43:53Z) - Uncertainty in Language Models: Assessment through Rank-Calibration [65.10149293133846]
Language Models (LMs) have shown promising performance in natural language generation.
It is crucial to correctly quantify their uncertainty in responding to given inputs.
We develop a novel and practical framework, termed $Rank$-$Calibration$, to assess uncertainty and confidence measures for LMs.
arXiv Detail & Related papers (2024-04-04T02:31:05Z) - SPUQ: Perturbation-Based Uncertainty Quantification for Large Language
Models [9.817185255633758]
Large language models (LLMs) have become increasingly prevalent, offering remarkable text generation capabilities.
A pressing challenge is their tendency to make confidently wrong predictions.
We introduce a novel UQ method, sampling with perturbation for UQ (SPUQ), designed to tackle both aleatoric and epistemic uncertainties.
Our findings show a substantial improvement in model calibration, with a reduction in Expected Error (ECE) by 50% on average.
arXiv Detail & Related papers (2024-03-04T21:55:22Z) - Decomposing Uncertainty for Large Language Models through Input Clarification Ensembling [69.83976050879318]
In large language models (LLMs), identifying sources of uncertainty is an important step toward improving reliability, trustworthiness, and interpretability.
In this paper, we introduce an uncertainty decomposition framework for LLMs, called input clarification ensembling.
Our approach generates a set of clarifications for the input, feeds them into an LLM, and ensembles the corresponding predictions.
arXiv Detail & Related papers (2023-11-15T05:58:35Z) - Quantifying Uncertainty in Natural Language Explanations of Large
Language Models [29.34960984639281]
Large Language Models (LLMs) are increasingly used as powerful tools for high-stakes natural language processing (NLP) applications.
We propose two novel metrics -- $textitVerbalized Uncertainty$ and $textitProbing Uncertainty$ -- to quantify the uncertainty of generated explanations.
Our empirical analysis of benchmark datasets reveals that verbalized uncertainty is not a reliable estimate of explanation confidence.
arXiv Detail & Related papers (2023-11-06T21:14:40Z)
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.