Modern neural language models achieve high accuracy in text generation, yet precise control over generation length remains underdeveloped. In this paper, we first investigate a recent length control method based on Reverse Positional Embeddings (RPE) and show its limits when control is requested beyond the training distribution. In particular, using a discrete countdown signal tied to the absolute remaining token count leads to instability. To provide robust length control, we introduce Progress Ratio Embeddings (PRE), as continuous embeddings tied to a trigonometric impatience signal. PRE integrates seamlessly into standard Transformer architectures, providing stable length fidelity without degrading text accuracy under standard evaluation metrics. We further show that PRE generalizes well to unseen target lengths. Experiments on two widely used news-summarization benchmarks validate these findings.

Transformer-based architectures have advanced text summarization, yet their quadratic complexity limits scalability on long documents. This paper introduces BiSparse-AAS (Bilinear Sparse Attention with Adaptive Spans), a novel framework that combines sparse attention, adaptive spans, and bilinear attention to address these limitations. Sparse attention reduces computational costs by focusing on the most relevant parts of the input, while adaptive spans dynamically adjust the attention ranges. Bilinear attention complements both by modeling complex token interactions within this refined context. BiSparse-AAS consistently outperforms state-of-the-art baselines in both extractive and abstractive summarization tasks, achieving average ROUGE improvements of about 68.1% on CNN/DailyMail and 52.6% on XSum, while maintaining strong performance on OpenWebText and Gigaword datasets. By addressing efficiency, scalability, and long-sequence modeling, BiSparse-AAS provides a unified, practical solution for real-world text summarization applications.

Recent advances in large language models (LLMs) have been driven by pretraining, supervised fine tuning (SFT), and alignment tuning. Among these, SFT plays a crucial role in transforming a model 's general knowledge into structured responses tailored to specific tasks. However, there is no clearly established methodology for effective training data selection. Simply increasing the volume of data does not guarantee performance improvements, while preprocessing, sampling, and validation require substantial time and cost. To address this issue, a variety of data selection methods have been proposed. Among them, knowledge based selection approaches identify suitable training data by analyzing the model 's responses. Nevertheless, these methods typically rely on prompt engineering, making them sensitive to variations and incurring additional costs for prompt design. In this study, we propose Knowledge Analysis via Model Internal Representations (KAMIR), a novel approach that overcomes these limitations by analyzing data based on the model 's internal representations. KAMIR computes similarities between the hidden states of each layer (block) and the final hidden states for a given input to assess the data. Unlike prior methods that were largely limited to multiple choice tasks, KAMIR can be applied to a wide range of tasks such as machine reading comprehension and summarization. Moreover, it selects data useful for training based on the model 's familiarity with the input, even with a small dataset and a simple classifier architecture. Experiments across diverse task datasets demonstrate that training with less familiar data leads to better generalization performance.

Controlling the length of generated text can be crucial in various text-generation tasks, including summarization. Existing methods often require complex model alterations, limiting compatibility with pre-trained models. We address these limitations by developing a simple approach for controlling the length of automatic text summaries by increasing the importance of correctly predicting the EOS token in the cross-entropy loss computation. The proposed methodology is agnostic to architecture and decoding algorithms and orthogonal to other inference-time techniques to control the generation length. We tested it with encoder-decoder and modern GPT-style LLMs, and show that this method can control generation length, often without affecting the quality of the summary.

The sentence-level framework defines extractive summarization as an individual sentence selection problem, determining whether each sentence in a document should be included in the summary. However, the sentence-level framework often produces summaries that contain only general sentences or repeat important but similar sentences (Narayan et al., 2018b; Zhong et al., 2020). The summary-level framework overcomes this limitation by defining extractive summarization as a summary ranking problem rather than a sentence selection problem. The main idea of the summary-level framework is to generate a set of candidate summaries consisting of different sentences, and then rank them to select the best summary. By considering sentence composition at the entire summary level rather than sentence by sentence, this approach enables each sentence in the summary to convey different, specific information (Narayan et al., 2018b; Zhong et al., 2020). Previous work in both frameworks has primarily focused on improving which sentences to include in the summary, or in other words, sentence inclusion. However, to the best of our knowledge, the importance of sentence order in summaries has not been highlighted since the era of graph-based extractive summarization (Mihalcea and Ta-rau, 2004; Erkan and Radev, 2004). The sentence order of a text plays a crucial role not only in readability but also in its meaning (Yin et al., 2019; Lo-geswaran et al., 2018). Table 1 illustrates how the arXiv:2502.16180v1

Text summarization tasks commonly employ Pre-trained Language Models (PLMs) to fit diverse standard datasets. While these PLMs excel in automatic evaluations, they frequently underperform in human evaluations, indicating a deviation between their generated summaries and human summarization preferences. This discrepancy is likely due to the low quality of fine-tuning datasets and the limited availability of high-quality human-annotated data that reflect true human preference. To address this challenge, we introduce a novel human summarization preference alignment framework AlignSum. This framework consists of three parts: Firstly, we construct a Data Pymarid with extractive, abstractive, and human-annotated summary data. Secondly, we conduct the Gaussian Resampling to remove summaries with extreme lengths. Finally, we implement the two-stage hierarchical fine-tuning with Data Pymarid after Gaussian Resampling. We apply AlignSum to PLMs on the human-annotated CNN/DailyMail and BBC XSum datasets. Experiments show that with AlignSum, PLMs like BART-Large surpass 175B GPT-3 in both automatic and human evaluations. This demonstrates that AlignSum significantly enhances the alignment of language models with human summarization preferences.

The extraction of essential news elements through the 5W1H framework (\textit{What}, \textit{When}, \textit{Where}, \textit{Why}, \textit{Who}, and \textit{How}) is critical for event extraction and text summarization. The advent of Large language models (LLMs) such as ChatGPT presents an opportunity to address language-related tasks through simple prompts without fine-tuning models with much time. While ChatGPT has encountered challenges in processing longer news texts and analyzing specific attributes in context, especially answering questions about \textit{What}, \textit{Why}, and \textit{How}. The effectiveness of extraction tasks is notably dependent on high-quality human-annotated datasets. However, the absence of such datasets for the 5W1H extraction increases the difficulty of fine-tuning strategies based on open-source LLMs. To address these limitations, first, we annotate a high-quality 5W1H dataset based on four typical news corpora (\textit{CNN/DailyMail}, \textit{XSum}, \textit{NYT}, \textit{RA-MDS}); second, we design several strategies from zero-shot/few-shot prompting to efficient fine-tuning to conduct 5W1H aspects extraction from the original news documents. The experimental results demonstrate that the performance of the fine-tuned models on our labelled dataset is superior to the performance of ChatGPT. Furthermore, we also explore the domain adaptation capability by testing the source-domain (e.g. NYT) models on the target domain corpus (e.g. CNN/DailyMail) for the task of 5W1H extraction.

Auto-regressive generation models achieve competitive performance across many different NLP tasks such as summarization, question answering, and classifications. However, they are also known for being slow in inference, which makes them challenging to deploy in real-time applications. We propose a switchable decision to accelerate inference by dynamically assigning computation resources for each data instance. Automatically making decisions on where to skip and how to balance quality and computation cost with constrained optimization, our dynamic neural generation networks enforce the efficient inference path and determine the optimized trade-off. Experiments across question answering, summarization, and classification benchmarks show that our method benefits from less computation cost during inference while keeping the same accuracy. Extensive experiments and ablation studies demonstrate that our method can be general, effective, and beneficial for many NLP tasks.

With the improvements in generative models, the issues of producing hallucinations in various domains (e.g., law, writing) have been brought to people's attention due to concerns about misinformation. In this paper, we focus on neural fake news, which refers to content generated by neural networks aiming to mimic the style of real news to deceive people. To prevent harmful disinformation spreading fallaciously from malicious social media (e.g., content farms), we propose a novel verification framework, Style-News, using publisher metadata to imply a publisher's template with the corresponding text types, political stance, and credibility. Based on threat modeling aspects, a style-aware neural news generator is introduced as an adversary for generating news content conditioning for a specific publisher, and style and source discriminators are trained to defend against this attack by identifying which publisher the style corresponds with, and discriminating whether the source of the given news is human-written or machine-generated. To evaluate the quality of the generated content, we integrate various dimensional metrics (language fluency, content preservation, and style adherence) and demonstrate that Style-News significantly outperforms the previous approaches by a margin of 0.35 for fluency, 15.24 for content, and 0.38 for style at most. Moreover, our discriminative model outperforms state-of-the-art baselines in terms of publisher prediction (up to 4.64%) and neural fake news detection (+6.94% $\sim$ 31.72%).

The current paradigm of large-scale pre-training and fine-tuning Transformer large language models has lead to significant improvements across the board in natural language processing. However, such large models are susceptible to overfitting to their training data, and as a result the models perform poorly when the domain changes. Also, due to the model's scale, the cost of fine-tuning the model to the new domain is large. Nonparametric Variational Information Bottleneck (NVIB) has been proposed as a regulariser for training cross-attention in Transformers, potentially addressing the overfitting problem. We extend the NVIB framework to replace all types of attention functions in Transformers, and show that existing pretrained Transformers can be reinterpreted as Nonparametric Variational (NV) models using a proposed identity initialisation. We then show that changing the initialisation introduces a novel, information-theoretic post-training regularisation in the attention mechanism, which improves out-of-domain generalisation without any training. This success supports the hypothesis that pretrained Transformers are implicitly NV Bayesian models.