Language models are strong few-shot learners and achieve good overall accuracy in text classification tasks, masking the fact that their results suffer from great class accuracy imbalance. We believe that the pursuit of overall accuracy should not come from enriching the strong classes, but from raising up the weak ones. To address the imbalance, we propose a post-hoc nonlinear integer programming based debiasing method that ensembles weight correction and membership correction to enable flexible rectifications of class probabilities at both class and sample levels, enhancing the performance of LLMs directly from their outputs. Evaluations with Llama-2-13B on seven text classification benchmarks show that our approach achieves state-of-the-art overall accuracy gains with balanced class accuracies. The resulted probability correction scheme demonstrates that sample-level corrections are necessary to elevate weak classes. In addition, due to effectively correcting weak classes, our method also brings significant performance gains to Llama-2-70B, especially on a biomedical domain task, demonstrating its effectiveness across both small and large model variants.

In-context learning, which allows large language models to perform diverse tasks with a few demonstrations, is found to have imbalanced per-class prediction accuracy on multi-class text classification. Although notable output correction methods have been developed to tackle the issue and simultaneously improve downstream prediction accuracy, they may fail to answer the core interpretability challenges: why and which certain classes need corrections, and more importantly, a tailored correction for per-sample, per-class's probability. To address such interpretability gaps, we first find that the imbalance arises from certain classes consistently receiving high ICL output probabilities, whereas others receiving lower or mixed ranges, so the former is more frequently chosen, resulting in higher accuracy; more crucially, we find that these ranges have significantly varying degrees of influence on the accuracy bias, highlighting the need for precise, interpretable probability corrections by range. Motivated by this, we propose FuRud, a Fuzzy Rule Optimization based Debiasing method, that (1) detects which classes need corrections, and (2) for each correction-needed class, detects its probability ranges and applies asymmetric amplifications or reductions to correct them interpretably. Notably, across seven benchmark datasets, FuRud reduces the pairwise class accuracy bias (COBias) by more than half (56%), while achieving a relative increase of 21% in accuracy, outperforming state-of-the-art debiasing methods. Moreover, FuRud can optimize downstream tasks with as few as 10 optimization examples. Furthermore, FuRud can work for prompt formats that lead to highly skewed predictions. For example, FuRud greatly improves ICL outputs which use letter options, with 44% relative accuracy increase and 54% relative COBias reduction.

Large language models (LLMs) have gained human trust due to their capabilities and helpfulness. However, this in turn may allow LLMs to affect users' mindsets by manipulating language. It is termed as gaslighting, a psychological effect. In this work, we aim to investigate the vulnerability of LLMs under prompt-based and fine-tuning-based gaslighting attacks. Therefore, we propose a two-stage framework DeepCoG designed to: 1) elicit gaslighting plans from LLMs with the proposed DeepGaslighting prompting template, and 2) acquire gaslighting conversations from LLMs through our Chain-of-Gaslighting method. The gaslighting conversation dataset along with a corresponding safe dataset is applied to fine-tuning-based attacks on open-source LLMs and anti-gaslighting safety alignment on these LLMs. Experiments demonstrate that both prompt-based and fine-tuning-based attacks transform three open-source LLMs into gaslighters. In contrast, we advanced three safety alignment strategies to strengthen (by 12.05%) the safety guardrail of LLMs. Our safety alignment strategies have minimal impacts on the utility of LLMs. Empirical studies indicate that an LLM may be a potential gaslighter, even if it passed the harmfulness test on general dangerous queries.

For language model classification, would you prefer having only one workable class or having every class working? The latter makes more practical uses. Especially for large language models (LLMs), the fact that they achieve a fair overall accuracy by in-context learning (ICL) obscures a large difference in individual class accuracies. In this work, we uncover and tackle language models' imbalance in per-class prediction accuracy by reconceptualizing it as the Contextual Oddity Bias (COBias), and we are the first to engage nonlinear integer programming (NIP) to debias it. Briefly, COBias refers to the difference in accuracy by a class A compared to its ''odd'' class, which holds the majority wrong predictions of class A. With the COBias metric, we reveal that LLMs of varied scales and families exhibit large per-class accuracy differences. Then we propose Debiasing as Nonlinear Integer Programming (DNIP) to correct ICL per-class probabilities for lower bias and higher overall accuracy. Our optimization objective is directly based on the evaluation scores by COBias and accuracy metrics, solved by simulated annealing. Evaluations on three LLMs across seven NLP classification tasks show that DNIP simultaneously achieves significant COBias reduction ($-27\%$) and accuracy improvement ($+12\%$) over the conventional ICL approach, suggesting that modeling pairwise class accuracy differences is a direction in pushing forward more accurate, more reliable LLM predictions.