Beyond Under-Alignment: Atomic Preference Enhanced Factuality Tuning for Large Language Models

• URL: http://arxiv.org/abs/2406.12416v2
• Date: Thu, 27 Jun 2024 12:07:55 GMT
• Title: Beyond Under-Alignment: Atomic Preference Enhanced Factuality Tuning for Large Language Models
• Authors: Hongbang Yuan, Yubo Chen, Pengfei Cao, Zhuoran Jin, Kang Liu, Jun Zhao,
• Abstract summary: We evaluate the factuality of different models tuned by various preference learning algorithms. We propose textbfAPEFT (textbfAtomic textbfPreference textbfEnhanced textbfFactuality textbfTuning) to enhance model's awareness of factuality.
• Score: 19.015202590038996
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large language models (LLMs) have achieved remarkable success but still tend to generate factually erroneous responses, a phenomenon known as hallucination. A recent trend is to use preference learning to fine-tune models to align with factuality. However, existing work primarily evaluates fine-tuned models on in-domain (ID) datasets and the factuality on out-of-domain (OOD) datasets remains underexplored. In this paper, we conduct a comprehensive evaluation of the factuality of different models tuned by various preference learning algorithms and demonstrate that their performance on OOD datasets either increases minimally or decreases. Subsequently, we reveal that the main cause of model's failure to uphold factuality under a distribution shift is \textbf{under-alignment}, rather than \textbf{over-alignment}, by analyzing the token distribution shift of the models before and after tuning. Finally, we propose \textbf{APEFT} (\textbf{A}tomic \textbf{P}reference \textbf{E}nhanced \textbf{F}actuality \textbf{T}uning), a framework that enhances model's awareness of factuality at the granularity of individual facts. Extensive experiments demonstrate that APEFT improves model performance by an average of $\boldsymbol{3.45\%}$ on both ID and OOD datasets, which is highly effective.

Related papers

• A Plug-and-Play Method for Rare Human-Object Interactions Detection by Bridging Domain Gap [50.079224604394]
  We present a novel model-agnostic framework called textbfContext-textbfEnhanced textbfFeature textbfAment (CEFA) CEFA consists of a feature alignment module and a context enhancement module. Our method can serve as a plug-and-play module to improve the detection performance of HOI models on rare categories.
  arXiv  Detail & Related papers  (2024-07-31T08:42:48Z)
• 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)
• Feature Protection For Out-of-distribution Generalization [24.072876186625855]
  We show that protecting pre-trained features leads to a fine-tuned model more robust to generalization. We show that protecting pre-trained features leads to a fine-tuned model more robust to OOD generalization.
  arXiv  Detail & Related papers  (2024-05-25T03:00:06Z)
• Entity-level Factual Adaptiveness of Fine-tuning based Abstractive Summarization Models [31.84120883461332]
  We analyze the robustness of fine-tuning based summarization models to the knowledge conflict. We introduce a controllable counterfactual data augmentation method.
  arXiv  Detail & Related papers  (2024-02-23T07:53:39Z)
• Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution [67.9215891673174]
  We propose score entropy as a novel loss that naturally extends score matching to discrete spaces. We test our Score Entropy Discrete Diffusion models on standard language modeling tasks.
  arXiv  Detail & Related papers  (2023-10-25T17:59:12Z)
• The Bayesian Context Trees State Space Model for time series modelling and forecasting [8.37609145576126]
  A hierarchical Bayesian framework is introduced for developing rich mixture models for real-valued time series. At the top level, meaningful discrete states are identified as appropriately quantised values of some of the most recent samples. At the bottom level, a different, arbitrary model for real-valued time series - a base model - is associated with each state.
  arXiv  Detail & Related papers  (2023-08-02T02:40:42Z)
• Preserving Knowledge Invariance: Rethinking Robustness Evaluation of Open Information Extraction [50.62245481416744]
  We present the first benchmark that simulates the evaluation of open information extraction models in the real world. We design and annotate a large-scale testbed in which each example is a knowledge-invariant clique. By further elaborating the robustness metric, a model is judged to be robust if its performance is consistently accurate on the overall cliques.
  arXiv  Detail & Related papers  (2023-05-23T12:05:09Z)
• Comparing Test Sets with Item Response Theory [53.755064720563]
  We evaluate 29 datasets using predictions from 18 pretrained Transformer models on individual test examples. We find that Quoref, HellaSwag, and MC-TACO are best suited for distinguishing among state-of-the-art models. We also observe span selection task format, which is used for QA datasets like QAMR or SQuAD2.0, is effective in differentiating between strong and weak models.
  arXiv  Detail & Related papers  (2021-06-01T22:33:53Z)
• Disfluency Detection with Unlabeled Data and Small BERT Models [3.04133054437883]
  We focus on the disfluency detection task, focusing on small, fast, on-device models based on the BERT architecture. We demonstrate it is possible to train disfluency detection models as small as 1.3 MiB, while retaining high performance.
  arXiv  Detail & Related papers  (2021-04-21T21:24:32Z)
• Elastic weight consolidation for better bias inoculation [24.12790037712358]
  Elastic weight consolidation (EWC) allows fine-tuning of models to mitigate biases. EWC dominates standard fine-tuning, yielding models with lower levels of forgetting on the original (biased) dataset.
  arXiv  Detail & Related papers  (2020-04-29T17:45:12Z)
• Decomposed Adversarial Learned Inference [118.27187231452852]
  We propose a novel approach, Decomposed Adversarial Learned Inference (DALI) DALI explicitly matches prior and conditional distributions in both data and code spaces. We validate the effectiveness of DALI on the MNIST, CIFAR-10, and CelebA datasets.
  arXiv  Detail & Related papers  (2020-04-21T20:00:35Z)

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.