In-context learning (ICL) has transformed the use of large language models (LLMs) for NLP tasks, enabling few-shot learning by conditioning on labeled examples without finetuning. Despite its effectiveness, ICL is prone to errors, especially for challenging examples. With the goal of improving the performance of ICL, we propose corrective in-context learning (CICL), an approach that incorporates a model's incorrect predictions alongside ground truth corrections into the prompt, aiming to enhance classification accuracy through self-correction. However, contrary to our hypothesis, extensive experiments on text classification tasks demonstrate that CICL consistently underperforms standard ICL, with performance degrading as the proportion of corrections in the prompt increases. Our findings indicate that CICL introduces confusion by disrupting the model's task understanding, rather than refining its predictions. Additionally, we observe that presenting harder examples in standard ICL does not improve performance, suggesting that example difficulty alone may not be a reliable criterion for effective selection. By presenting these negative results, we provide important insights into the limitations of self-corrective mechanisms in LLMs and offer directions for future research.
Text classification, a fundamental task in natural language processing, aims to categorize textual data into predefined labels. Traditional methods struggled with complex linguistic structures and semantic dependencies. However, the advent of deep learning, particularly recurrent neural networks and Transformer-based models, has significantly advanced the field by enabling nuanced feature extraction and context-aware predictions. Despite these improvements, existing models still exhibit limitations in balancing interpretability, computational efficiency, and long-range contextual understanding. To address these challenges, this paper proposes the Dynamic Bidirectional Elman with Attention Network (DBEAN). DBEAN integrates bidirectional temporal modeling with self-attention mechanisms. It dynamically assigns weights to critical segments of input, improving contextual representation while maintaining computational efficiency.
In real-world text classification tasks, negative texts often contain a minimal proportion of negative content, which is especially problematic in areas like text quality control, legal risk screening, and sensitive information interception. This challenge manifests at two levels: at the macro level, distinguishing negative texts is difficult due to the high similarity between coarse-grained positive and negative samples; at the micro level, the issue stems from extreme class imbalance and a lack of fine-grained labels. To address these challenges, we propose transforming the coarse-grained positive-negative (PN) classification task into an imbalanced fine-grained positive-unlabeled (PU) classification problem, supported by theoretical analysis. We introduce a novel framework, Balanced Fine-Grained Positive-Unlabeled (BFGPU) learning, which features a unique PU learning loss function that optimizes macro-level performance amidst severe imbalance at the micro level. The framework's performance is further boosted by rebalanced pseudo-labeling and threshold adjustment. Extensive experiments on both public and real-world datasets demonstrate the effectiveness of BFGPU, which outperforms other methods, even in extreme scenarios where both macro and micro levels are highly imbalanced.
Large Language Models (LLMs) have demonstrated potential in predicting mental health outcomes from online text, yet traditional classification methods often lack interpretability and robustness. This study evaluates structured reasoning techniques-Chain-of-Thought (CoT), Self-Consistency (SC-CoT), and Tree-of-Thought (ToT)-to improve classification accuracy across multiple mental health datasets sourced from Reddit. We analyze reasoning-driven prompting strategies, including Zero-shot CoT and Few-shot CoT, using key performance metrics such as Balanced Accuracy, F1 score, and Sensitivity/Specificity. Our findings indicate that reasoning-enhanced techniques improve classification performance over direct prediction, particularly in complex cases. Compared to baselines such as Zero Shot non-CoT Prompting, and fine-tuned pre-trained transformers such as BERT and Mental-RoBerta, and fine-tuned Open Source LLMs such as Mental Alpaca and Mental-Flan-T5, reasoning-driven LLMs yield notable gains on datasets like Dreaddit (+0.52\% over M-LLM, +0.82\% over BERT) and SDCNL (+4.67\% over M-LLM, +2.17\% over BERT). However, performance declines in Depression Severity, and CSSRS predictions suggest dataset-specific limitations, likely due to our using a more extensive test set. Among prompting strategies, Few-shot CoT consistently outperforms others, reinforcing the effectiveness of reasoning-driven LLMs. Nonetheless, dataset variability highlights challenges in model reliability and interpretability. This study provides a comprehensive benchmark of reasoning-based LLM techniques for mental health text classification. It offers insights into their potential for scalable clinical applications while identifying key challenges for future improvements.
Extracting a small subset of crucial rationales from the full input is a key problem in explainability research. The most widely used fundamental criterion for rationale extraction is the maximum mutual information (MMI) criterion. In this paper, we first demonstrate that MMI suffers from diminishing marginal returns. Once part of the rationale has been identified, finding the remaining portions contributes only marginally to increasing the mutual information, making it difficult to use MMI to locate the rest. In contrast to MMI that aims to reproduce the prediction, we seek to identify the parts of the input that the network can actually utilize. This is achieved by comparing how different rationale candidates match the capability space of the weight matrix. The weight matrix of a neural network is typically low-rank, meaning that the linear combinations of its column vectors can only cover part of the directions in a high-dimensional space (high-dimension: the dimensions of an input vector). If an input is fully utilized by the network, {it generally matches these directions (e.g., a portion of a hypersphere), resulting in a representation with a high norm. Conversely, if an input primarily falls outside (orthogonal to) these directions}, its representation norm will approach zero, behaving like noise that the network cannot effectively utilize. Building on this, we propose using the norms of rationale candidates as an alternative objective to MMI. Through experiments on four text classification datasets and one graph classification dataset using three network architectures (GRUs, BERT, and GCN), we show that our method outperforms MMI and its improved variants in identifying better rationales. We also compare our method with a representative LLM (llama-3.1-8b-instruct) and find that our simple method gets comparable results to it and can sometimes even outperform it.
Financial Sentiment Analysis (FSA) traditionally relies on human-annotated sentiment labels to infer investor sentiment and forecast market movements. However, inferring the potential market impact of words based on their human-perceived intentions is inherently challenging. We hypothesize that the historical market reactions to words, offer a more reliable indicator of their potential impact on markets than subjective sentiment interpretations by human annotators. To test this hypothesis, a market-derived labeling approach is proposed to assign tweet labels based on ensuing short-term price trends, enabling the language model to capture the relationship between textual signals and market dynamics directly. A domain-specific language model was fine-tuned on these labels, achieving up to an 11% improvement in short-term trend prediction accuracy over traditional sentiment-based benchmarks. Moreover, by incorporating market and temporal context through prompt-tuning, the proposed context-aware language model demonstrated an accuracy of 89.6% on a curated dataset of 227 impactful Bitcoin-related news events with significant market impacts. Aggregating daily tweet predictions into trading signals, our method outperformed traditional fusion models (which combine sentiment-based and price-based predictions). It challenged the assumption that sentiment-based signals are inferior to price-based predictions in forecasting market movements. Backtesting these signals across three distinct market regimes yielded robust Sharpe ratios of up to 5.07 in trending markets and 3.73 in neutral markets. Our findings demonstrate that language models can serve as effective short-term market predictors. This paradigm shift underscores the untapped capabilities of language models in financial decision-making and opens new avenues for market prediction applications.
Conformal prediction is a distribution-free framework for uncertainty quantification that replaces point predictions with sets, offering marginal coverage guarantees (i.e., ensuring that the prediction sets contain the true label with a specified probability, in expectation). In this paper, we uncover a novel connection between conformal prediction and sparse softmax-like transformations, such as sparsemax and $\gamma$-entmax (with $\gamma > 1$), which may assign nonzero probability only to a subset of labels. We introduce new non-conformity scores for classification that make the calibration process correspond to the widely used temperature scaling method. At test time, applying these sparse transformations with the calibrated temperature leads to a support set (i.e., the set of labels with nonzero probability) that automatically inherits the coverage guarantees of conformal prediction. Through experiments on computer vision and text classification benchmarks, we demonstrate that the proposed method achieves competitive results in terms of coverage, efficiency, and adaptiveness compared to standard non-conformity scores based on softmax.
Uncertainty quantification (UQ) is a prominent approach for eliciting truthful answers from large language models (LLMs). To date, information-based and consistency-based UQ have been the dominant UQ methods for text generation via LLMs. Density-based methods, despite being very effective for UQ in text classification with encoder-based models, have not been very successful with generative LLMs. In this work, we adapt Mahalanobis Distance (MD) - a well-established UQ technique in classification tasks - for text generation and introduce a new supervised UQ method. Our method extracts token embeddings from multiple layers of LLMs, computes MD scores for each token, and uses linear regression trained on these features to provide robust uncertainty scores. Through extensive experiments on eleven datasets, we demonstrate that our approach substantially improves over existing UQ methods, providing accurate and computationally efficient uncertainty scores for both sequence-level selective generation and claim-level fact-checking tasks. Our method also exhibits strong generalization to out-of-domain data, making it suitable for a wide range of LLM-based applications.
Masked language modeling has become a widely adopted unsupervised technique to pre-train language models. However, the process of selecting tokens for masking is random, and the percentage of masked tokens is typically fixed for the entire training process. In this paper, we propose to adjust the masking ratio and to decide which tokens to mask based on a novel task-informed anti-curriculum learning scheme. First, we harness task-specific knowledge about useful and harmful tokens in order to determine which tokens to mask. Second, we propose a cyclic decaying masking ratio, which corresponds to an anti-curriculum schedule (from hard to easy). We exemplify our novel task-informed anti-curriculum by masking (TIACBM) approach across three diverse downstream tasks: sentiment analysis, text classification by topic, and authorship attribution. Our findings suggest that TIACBM enhances the ability of the model to focus on key task-relevant features, contributing to statistically significant performance gains across tasks. We release our code at https://github.com/JarcaAndrei/TIACBM.
Autoregressive Language Models output text by sequentially predicting the next token to generate, with modern methods like Chain-of-Thought (CoT) prompting achieving state-of-the-art reasoning capabilities by scaling the number of generated tokens. However, are there times when we can infer how the model will behave (e.g. abstain from answering a question) early in the computation, making generation unnecessary? We show that internal representation of input tokens alone can often precisely predict, not just the next token, but eventual behavior over the entire output sequence. We leverage this capacity and learn probes on internal states to create early warning (and exit) systems. Specifically, if the probes can confidently estimate the way the LM is going to behave, then the system will avoid generating tokens altogether and return the estimated behavior instead. On 27 text classification datasets spanning five different tasks, we apply this method to estimate the eventual answer of an LM under CoT prompting, reducing inference costs by 65% (average) while suffering an accuracy loss of no more than 1.4% (worst case). We demonstrate the potential of this method to pre-emptively identify when a model will abstain from answering a question, fail to follow output format specifications, or give a low-confidence response. We explore the limits of this capability, showing that probes generalize to unseen datasets, but perform worse when LM outputs are longer and struggle to predict properties that require access to knowledge that the models themselves lack. Encouragingly, performance scales with model size, suggesting applicability to the largest of models
The computational cost of softmax-based attention in transformers limits their applicability to long-context tasks. Adaptive sparsity, of which $\alpha$-entmax attention is an example, offers a flexible data-dependent alternative, but existing implementations are inefficient and do not leverage the sparsity to obtain runtime and memory gains. In this work, we propose AdaSplash, which combines the efficiency of GPU-optimized algorithms with the sparsity benefits of $\alpha$-entmax. We first introduce a hybrid Halley-bisection algorithm, resulting in a 7-fold reduction in the number of iterations needed to compute the $\alpha$-entmax transformation. Then, we implement custom Triton kernels to efficiently handle adaptive sparsity. Experiments with RoBERTa and ModernBERT for text classification and single-vector retrieval, along with GPT-2 for language modeling, show that our method achieves substantial improvements in runtime and memory efficiency compared to existing $\alpha$-entmax implementations. It approaches -- and in some cases surpasses -- the efficiency of highly optimized softmax implementations like FlashAttention-2, enabling long-context training while maintaining strong task performance.
Large language models (LLMs) are widely adopted to generate synthetic datasets for various natural language processing (NLP) tasks, such as text classification and summarization. However, accurately measuring the diversity of these synthetic datasets-an aspect crucial for robust model performance-remains a significant challenge. In this paper, we introduce DCScore, a novel method for measuring synthetic dataset diversity from a classification perspective. Specifically, DCScore formulates diversity evaluation as a sample classification task, leveraging mutual relationships among samples. We further provide theoretical verification of the diversity-related axioms satisfied by DCScore, highlighting its role as a principled diversity evaluation method. Experimental results on synthetic datasets reveal that DCScore enjoys a stronger correlation with multiple diversity pseudo-truths of evaluated datasets, underscoring its effectiveness. Moreover, both empirical and theoretical evidence demonstrate that DCScore substantially reduces computational costs compared to existing approaches. Code is available at: https://github.com/BlueWhaleLab/DCScore.
Large language models (LLMs) have facilitated the generation of high-quality, cost-effective synthetic data for developing downstream models and conducting statistical analyses in various domains. However, the increased reliance on synthetic data may pose potential negative impacts. Numerous studies have demonstrated that LLM-generated synthetic data can perpetuate and even amplify societal biases and stereotypes, and produce erroneous outputs known as ``hallucinations'' that deviate from factual knowledge. In this paper, we aim to audit artifacts, such as classifiers, generators, or statistical plots, to identify those trained on or derived from synthetic data and raise user awareness, thereby reducing unexpected consequences and risks in downstream applications. To this end, we take the first step to introduce synthetic artifact auditing to assess whether a given artifact is derived from LLM-generated synthetic data. We then propose an auditing framework with three methods including metric-based auditing, tuning-based auditing, and classification-based auditing. These methods operate without requiring the artifact owner to disclose proprietary training details. We evaluate our auditing framework on three text classification tasks, two text summarization tasks, and two data visualization tasks across three training scenarios. Our evaluation demonstrates the effectiveness of all proposed auditing methods across all these tasks. For instance, black-box metric-based auditing can achieve an average accuracy of $0.868 \pm 0.071$ for auditing classifiers and $0.880 \pm 0.052$ for auditing generators using only 200 random queries across three scenarios. We hope our research will enhance model transparency and regulatory compliance, ensuring the ethical and responsible use of synthetic data.
Recent approaches in hierarchical text classification (HTC) rely on the capabilities of a pre-trained transformer model and exploit the label semantics and a graph encoder for the label hierarchy. In this paper, we introduce an effective hierarchical text classifier RADAr (Transformer-based Autoregressive Decoder Architecture) that is based only on an off-the-shelf RoBERTa transformer to process the input and a custom autoregressive decoder with two decoder layers for generating the classification output. Thus, unlike existing approaches for HTC, the encoder of RADAr has no explicit encoding of the label hierarchy and the decoder solely relies on the label sequences of the samples observed during training. We demonstrate on three benchmark datasets that RADAr achieves results competitive to the state of the art with less training and inference time. Our model consistently performs better when organizing the label sequences from children to parents versus the inverse, as done in existing HTC approaches. Our experiments show that neither the label semantics nor an explicit graph encoder for the hierarchy is needed. This has strong practical implications for HTC as the architecture has fewer requirements and provides a speed-up by a factor of 2 at inference time. Moreover, training a separate decoder from scratch in conjunction with fine-tuning the encoder allows future researchers and practitioners to exchange the encoder part as new models arise. The source code is available at https://github.com/yousef-younes/RADAr.
Attention mechanism is contributing to the majority of recent advances in machine learning for natural language processing. Additionally, it results in an attention map that shows the proportional influence of each input in its decision. Empirical studies postulate that attention maps can be provided as an explanation for model output. However, it is still questionable to ask whether this explanation helps regular people to understand and accept the model output (the plausibility of the explanation). Recent studies show that attention weights in the RNN encoders are hardly plausible because they spread on input tokens. We thus propose 3 additional constraints to the learning objective function to improve the plausibility of the attention map: regularization to increase the attention weight sparsity, semi-supervision to supervise the map by a heuristic and supervision by human annotation. Results show that all techniques can improve the attention map plausibility at some level. We also observe that specific instructions for human annotation might have a negative effect on classification performance. Beyond the attention map, the result of experiments on text classification tasks also shows that no matter how the constraint brings the gain, the contextualization layer plays a crucial role in finding the right space for finding plausible tokens.
This research introduces a hybrid classical-quantum framework for text classification, integrating GPT-Neo 125M with Low-Rank Adaptation (LoRA) and Synthetic Minority Over-sampling Technique (SMOTE) using quantum computing backends. While the GPT-Neo 125M baseline remains the best-performing model, the implementation of LoRA and SMOTE enhances the hybrid model, resulting in improved accuracy, faster convergence, and better generalization. Experiments on IBM's 127-qubit quantum backend and Pennylane's 32-qubit simulation demonstrate the viability of combining classical neural networks with quantum circuits. This framework underscores the potential of hybrid architectures for advancing natural language processing applications.
Short text classification has gained significant attention in the information age due to its prevalence and real-world applications. Recent advancements in graph learning combined with contrastive learning have shown promising results in addressing the challenges of semantic sparsity and limited labeled data in short text classification. However, existing models have certain limitations. They rely on explicit data augmentation techniques to generate contrastive views, resulting in semantic corruption and noise. Additionally, these models only focus on learning the intrinsic consistency between the generated views, neglecting valuable discriminative information from other potential views. To address these issues, we propose a Simple graph contrastive learning framework for Short Text Classification (SimSTC). Our approach involves performing graph learning on multiple text-related component graphs to obtain multi-view text embeddings. Subsequently, we directly apply contrastive learning on these embeddings. Notably, our method eliminates the need for data augmentation operations to generate contrastive views while still leveraging the benefits of multi-view contrastive learning. Despite its simplicity, our model achieves outstanding performance, surpassing large language models on various datasets.
Short text classification, as a research subtopic in natural language processing, is more challenging due to its semantic sparsity and insufficient labeled samples in practical scenarios. We propose a novel model named MI-DELIGHT for short text classification in this work. Specifically, it first performs multi-source information (i.e., statistical information, linguistic information, and factual information) exploration to alleviate the sparsity issues. Then, the graph learning approach is adopted to learn the representation of short texts, which are presented in graph forms. Moreover, we introduce a dual-level (i.e., instance-level and cluster-level) contrastive learning auxiliary task to effectively capture different-grained contrastive information within massive unlabeled data. Meanwhile, previous models merely perform the main task and auxiliary tasks in parallel, without considering the relationship among tasks. Therefore, we introduce a hierarchical architecture to explicitly model the correlations between tasks. We conduct extensive experiments across various benchmark datasets, demonstrating that MI-DELIGHT significantly surpasses previous competitive models. It even outperforms popular large language models on several datasets.
This study presents a comprehensive review of the potential of multimodal deep learning (DL) in medical diagnosis, using COVID-19 as a case example. Motivated by the success of artificial intelligence applications during the COVID-19 pandemic, this research aims to uncover the capabilities of DL in disease screening, prediction, and classification, and to derive insights that enhance the resilience, sustainability, and inclusiveness of science, technology, and innovation systems. Adopting a systematic approach, we investigate the fundamental methodologies, data sources, preprocessing steps, and challenges encountered in various studies and implementations. We explore the architecture of deep learning models, emphasising their data-specific structures and underlying algorithms. Subsequently, we compare different deep learning strategies utilised in COVID-19 analysis, evaluating them based on methodology, data, performance, and prerequisites for future research. By examining diverse data types and diagnostic modalities, this research contributes to scientific understanding and knowledge of the multimodal application of DL and its effectiveness in diagnosis. We have implemented and analysed 11 deep learning models using COVID-19 image, text, and speech (ie, cough) data. Our analysis revealed that the MobileNet model achieved the highest accuracy of 99.97% for COVID-19 image data and 93.73% for speech data (i.e., cough). However, the BiGRU model demonstrated superior performance in COVID-19 text classification with an accuracy of 99.89%. The broader implications of this research suggest potential benefits for other domains and disciplines that could leverage deep learning techniques for image, text, and speech analysis.
This paper studies the problem of stochastic bilevel optimization where the upper-level function is nonconvex with potentially unbounded smoothness and the lower-level function is strongly convex. This problem is motivated by meta-learning applied to sequential data, such as text classification using recurrent neural networks, where the smoothness constant of the upper-level loss function scales linearly with the gradient norm and can be potentially unbounded. Existing algorithm crucially relies on the nested loop design, which requires significant tuning efforts and is not practical. In this paper, we address this issue by proposing a Single Loop bIlevel oPtimizer (SLIP). The proposed algorithm first updates the lower-level variable by a few steps of stochastic gradient descent, and then simultaneously updates the upper-level variable by normalized stochastic gradient descent with momentum and the lower-level variable by stochastic gradient descent. Under standard assumptions, we show that our algorithm finds an $\epsilon$-stationary point within $\widetilde{O}(1/\epsilon^4)$\footnote{Here $\widetilde{O}(\cdot)$ compresses logarithmic factors of $1/\epsilon$ and $1/\delta$, where $\delta\in(0,1)$ denotes the failure probability.} oracle calls of stochastic gradient or Hessian-vector product, both in expectation and with high probability. This complexity result is nearly optimal up to logarithmic factors without mean-square smoothness of the stochastic gradient oracle. Our proof relies on (i) a refined characterization and control of the lower-level variable and (ii) establishing a novel connection between bilevel optimization and stochastic optimization under distributional drift. Our experiments on various tasks show that our algorithm significantly outperforms strong baselines in bilevel optimization.