Long prompt leads to huge hardware costs when using transformer-based Large Language Models (LLMs). Unfortunately, many tasks, such as summarization, inevitably introduce long documents, and the wide application of in-context learning easily ...

• We are the first to use the frozen LLM itself to compress over-limit prompts.
• We achieve a balance among training cost, inference efficiency, and response quality.
• Our method is more general and cost-efficient than existing ...

Deep learning has driven research in object detection and achieved proud results. Despite its significant advancements in object detection, small object detection still struggles with low recognition rates and inaccurate positioning, primarily ...

• Propose a method named DFFN to boost the detection precision of small objects.
• Introduce a VPEM module that renders neural networks perceptually adaptable.
• Incorporate class and bounding box alignment parts to facilitate the ...

Despite recent advances in large language models (LLMs), the best effectiveness in information extraction (IE) is still achieved by fine-tuned models, hence the need for manually annotated datasets to train them. However, collecting human ...

B-mode ultrasound (BUS) mainly reflects the tissue structural, morphological, and echo characteristics of liver tumors, and contrast-enhanced ultrasound (CEUS) offers supplementary information on the dynamic blood perfusion pattern to promote ...

Table-based fact verification focuses on determining the truthfulness of statements by cross-referencing data in tables. This task is challenging due to the complex interactions inherent in table structures. To address this challenge, existing ...

Display Omitted

• We propose Macol, which verifies statement accuracy by integrating relevant tables.
• By fusing multi-aspect reasoning clues, Macol guides us to obtain key insights.
• Using auto-generated adversarial instances, Macol enables fine-...

Continuous sign language recognition (CSLR) is essential for the social participation of deaf individuals. The structural information of sign language motion units plays a crucial role in semantic representation. However, most existing CSLR ...

Previous temporal knowledge graph (TKG) reasoning methods often focus exclusively on evolving representations. However, these methods suffer from the inadequacy of capturing the structural nuances of concurrent facts, the intricate relations in ...

• Integrating topological relation graphs and temporal fusion information.
• Time-aware relation attention mechanism captures structural dependencies.
• Designing topological patterns to generate topological relation graphs.
• ...

• We introduce ChatGPT paraphrased reviews as a new challenge for online fake reviews.
• We study the association/disassociation between real and paraphrased reviews.
• We identify a consistent association/disassociation pattern.
• ...

Fake reviews corrode the trust between businesses and consumers and distort the online image of a service or a product. The problem of fake review contamination is only going to worsen with the introduction of Artificial Intelligence (AI) ...

Multi-source cross-lingual transfer aims to acquire task knowledge from multiple labelled source languages and transfer it to an unlabelled target language, enabling effective performance in this target language. The existing methods mainly focus ...

• Propose a discrepancy-guided de-noising method to learn discriminative representations.
• Propose a pseudo-label supervised mutual learning method to promote mutual guidance.
• Propose a de-noising mutual learning method for multi-...

Depression detection based on text analysis has emerged as a research hotspot. Existing research indicates that patients’ personalized characteristics are the primary factor contributing to differences in reported experiences, which poses ...

• Pioneered personalized modeling in text-based depression detection.
• Personalized models narrow the gap between generic symptoms and patient experiences.
• Defined key components of personalized information and proposed a novel ...

Empathetic response generation endeavours to perceive the interlocutor’s emotional and cognitive states in the dialogue and express proper responses. Previous studies detect the interlocutor’s states by understanding the immediate context of the ...

• We introduce situations and explore their explicit and implicit associations.
• We propose a bidirectional filtering encoder to captures the explicit associations.
• We propose a reasoning knowledge-based hypergraph neural network.

Fake news detection involves developing techniques to identify and flag misleading or false information disseminated through media sources. Current efforts often use limited information for categorization, lacking comprehensive data integration ...

• We propose a novel Multifaceted Reasoning Network for Explainable Fake News Detection.
• We integrate multi-source and multi-granularity information, and provide multi-angle explanations for fake news classification.
• We conducted ...

Large language models (LLMs) that exhibit a remarkable ability by in-context learning (ICL) with bilingual demonstrations have been recognized as a potential solution for machine translation. However, the process of selecting these demonstrations ...

• We explore how to enhance translation ability and efficiency of large language model.
• A new product quantization technique to accelerate selecting demonstrations.
• An innovative template design for in-context learning to implement ...

The objective of the Aspect Sentiment Triplet Extraction (ASTE) task is to identify triplets of (aspect, opinion, sentiment) from user-generated reviews. The current study does not extensively integrate the interaction between word pairs and ...

• We develop SBRS for ASTE to effectively address the problem of overlapping triplets.
• SBRS designs two progressive word levels and word pair levels to refine the text.
• SBRS utilize transfer of temporal semantic information through ...

This study presents a financial distress prediction model focusing on the linguistic analysis of risk-related sections of corporate annual reports. Here, we introduce a novel methodology that leverages BERT-based contextualized embedding models ...

• Linguistic analysis of risk-related sections of corporate annual reports.
• BERT-based models used to identify financial sentiment and coherent topics.
• A semi-supervised XGBoost approach used to predict financial distress.
• ...

Person Re-identification (ReID) could search for the same pedestrian from non-overlapping cameras, which completes the pedestrian location and search purpose. However, the process contains much sensitive pedestrian information and raises serious ...

• We are the first to study the effective model switch to control the system status.
• We develop the universal adversarial algorithm to produce an ingenious start trigger.
• We design a model-aware training strategy to perceive issued ...

• Rule knowledge graph (KG) utilizes multi-dimensional semantic associations among rules to alleviate semantic drift in rule discovery.
• Label-aware rule generation approach realizes attentive semantic information propagation based on ...

In weakly supervised learning, labeling rules can automatically label data to train models. However, due to insufficient prior knowledge, rule discovery often suffers from semantic drift. Since misclassified rules are generated from wrongly ...

Stance detection, a fundamental task in natural language processing, identifies user stances in texts towards specific targets. The diverse targets and ever-changing expressions make it challenging to attain comprehensive knowledge from limited ...

• Target-aware knowledge is essential for stance detection in few-shot scenarios.
• Consistency between prior and stance knowledge ensures the power of prompt-tuning.
• The method uses a two-stage framework to ensure fusion knowledge ...

In the finance domain, nested named entities recognition has become a hot topic in named entity recognition tasks. Traditional nested entity recognition methods easily ignore the dependency relationships between entities, and these methods are ...

The task of event relation extraction (ERE) aims to organize multiple events and their relations as a directed graph. However, existing ERE methods exhibit two limitations: (1) Events in a document are typically expressed with merely a trigger ...

• We propose a graph propagation model for event relation extraction.
• We investigate rich event structures for this task with a novel joint model.
• A novel triadic contrastive training method to enable high-order interactions.
• It ...