Detecting Alzheimer’s disease (AD) from narrative transcripts challenges large language models (LLMs): pre-training rarely covers this out-of-distribution task, and all transcript demos describe the same scene, producing highly homogeneous contexts. These factors cripple both the model’s built-in task knowledge (task cognition) and its ability to surface subtle, class-discriminative cues (contextual perception). Because cognition is fixed after pre-training, improving in-context learning (ICL) for AD detection hinges on enriching perception through better demonstration (demo) sets. We demonstrate that standard ICL quickly saturates, its demos lack diversity (context width) and fail to convey fine-grained signals (context depth), and that recent task vector (TV) approaches improve broad task adaptation by injecting TV into the LLMs' hidden states (HSs), they are ill-suited for AD detection due to the mismatch of injection granularity, strength and position. To address these bottlenecks, we introduce DA4ICL, a demo-centric anchoring framework that jointly expands context width via Diverse and Contrastive Retrieval (DCR) and deepens each demo's signal via Projected Vector Anchoring (PVA) at every Transformer layer. Across three AD benchmarks, DA4ICL achieves large, stable gains over both ICL and TV baselines, charting a new paradigm for fine-grained, OOD and low-resource LLM adaptation.

Visual Question Answering (VQA) has increasingly shifted from direct prediction to step-by-step reasoning, driven by advances in Multi-modal Large Language Models (MLLMs) and Multi-modal Chain-of-Thought (MCoT). However, small MLLMs (S-MLLMs) often suffer from a "Faithfulness Gap", where they arrive at correct answers with inconsistent or hallucinatory intermediate reasoning processes. In this work, we first conduct a systematic diagnostic of MCoT consistency and identify two prevalent patterns that undermine conventional trajectory-level optimization: Intermediate Error and Inconsistent Reasoning. We categorize intermediate errors into Caption Errors—Hallucination, Contradiction, and Irrelevance, and Reasoning Errors—Divergence and Spuriousness. Driven by these findings, we curate MCoT-SP, a Stage-wise Preference dataset with severity-aware annotations constructed through an automated pipeline that detects and corrects intermediate errors using entailment-based checks and LLM-based correction, followed by post-verification to ensure the reliability of the preference pairs. Leveraging this resource, we propose Intermediate Process Direct Preference Optimization (IP-DPO), which performs stage-wise preference optimization by treating each intermediate stage as independent preference units, while jointly integrating preference learning with supervised fine-tuning on preferred complete MCoT responses to preserve global coherence. Extensive evaluations on LLaVA-Next and Qwen2.5-VL demonstrate that IP-DPO outperforms standard DPO and SFT baselines across two benchmarks spanning six VQA datasets. Our analysis indicates that improvements in intermediate consistency substantially exceed the corresponding improvements in final answer accuracy, revealing that IP-DPO primarily improves the underlying reasoning process rather than merely overfitting to target labels.

BACKGROUND: The potential of Segment Anything Model 2 (SAM2) for 3D medical image segmentation via video-stream processing is currently constrained by its reliance on manual prompts. While existing research employs auxiliary models (e.g., YOLO) as prompt generators, these approaches face two fundamental limitations: the inherent bottleneck of external models' feature extraction and the lack of mechanisms to prevent the propagation of erroneous prompts. Furthermore, current methods often struggle with interference from non-salient regions in complex 3D tumor datasets. This study aims to develop an automated, reliable prompt generation and sequence processing framework specifically for 3D medical imaging. RESULTS: We propose AutoPrompt-SAM3D, featuring an Automatic Prompt Generator that hierarchically integrates SAM2's tri-layer features and a supervised confidence frames filter for reliable prompt selection. Additionally, we implement a full-sequence processing framework that progressively localizes salient regions across consecutive slices. Comprehensive experiments conducted on four public abdominal tumor datasets demonstrate that AutoPrompt-SAM3D achieves superior 3D medical segmentation performance, consistently outperforming or matching state-of-the-art prompt-based methods. CONCLUSIONS: AutoPrompt-SAM3D eliminates the dependency on manual prompts in SAM2-based 3D segmentation through hierarchical feature integration and error filtering. By enhancing both the reliability and efficiency of tumor localization, this framework provides a practical tool for large-scale medical image analysis and supports more consistent clinical decision-making.

Detecting Alzheimer's Disease (AD) from narrative transcripts remains a challenging task for large language models (LLMs), particularly under out-of-distribution (OOD) and datascarce conditions. While in-context learning (ICL) provides a parameter-efficient alternative to fine-tuning, existing ICL approaches often suffer from task recognition failure, suboptimal demonstration selection, and misalignment between label words and task objectives, issues that are amplified in clinical domains like AD detection. We propose Explicit Knowledge In-Context Learners (EK-ICL), a novel framework that integrates structured explicit knowledge to enhance reasoning stability and task alignment in ICL. EK-ICL incorporates three knowledge components: confidence scores derived from small language models (SLMs) to ground predictions in task-relevant patterns, parsing feature scores to capture structural differences and improve demo selection, and label word replacement to resolve semantic misalignment with LLM priors. In addition, EK-ICL employs a parsing-based retrieval strategy and ensemble prediction to mitigate the effects of semantic homogeneity in AD transcripts. Extensive experiments across three AD datasets demonstrate that EK-ICL significantly outperforms state-of-the-art fine-tuning and ICL baselines. Further analysis reveals that ICL performance in AD detection is highly sensitive to the alignment of label semantics and task-specific context, underscoring the importance of explicit knowledge in clinical reasoning under low-resource conditions.

Recent advancements have successfully harnessed the power of Large Language Models (LLMs) for zero-shot document ranking, exploring a variety of prompting strategies. Comparative approaches like pairwise and listwise achieve high effectiveness but are computationally intensive and thus less practical for larger-scale applications. Scoring-based pointwise approaches exhibit superior efficiency by independently and simultaneously generating the relevance scores for each candidate document. However, this independence ignores critical comparative insights between documents, resulting in inconsistent scoring and suboptimal performance. In this paper, we aim to improve the effectiveness of pointwise methods while preserving their efficiency through two key innovations: (1) We propose a novel Global-Consistent Comparative Pointwise Ranking (GCCP) strategy that incorporates global reference comparisons between each candidate and an anchor document to generate contrastive relevance scores. We strategically design the anchor document as a query-focused summary of pseudo-relevant candidates, which serves as an effective reference point by capturing the global context for document comparison. (2) These contrastive relevance scores can be efficiently Post-Aggregated with existing pointwise methods, seamlessly integrating essential Global Context information in a training-free manner (PAGC). Extensive experiments on the TREC DL and BEIR benchmark demonstrate that our approach significantly outperforms previous pointwise methods while maintaining comparable efficiency. Our method also achieves competitive performance against comparative methods that require substantially more computational resources. More analyses further validate the efficacy of our anchor construction strategy.

The Segment Anything Model (SAM) has recently demonstrated significant potential in medical image segmentation. Although SAM is primarily trained on 2D images, attempts have been made to apply it to 3D medical image segmentation. However, the pseudo 3D processing used to adapt SAM results in spatial feature loss, limiting its performance. Additionally, most SAM-based methods still rely on manual prompts, which are challenging to implement in real-world scenarios and require extensive external expert knowledge. To address these limitations, we introduce the Decoder Enhanced and Auto Prompt SAM (DEAP-3DSAM) to tackle these limitations. Specifically, we propose a Feature Enhanced Decoder that fuses the original image features with rich and detailed spatial information to enhance spatial features. We also design a Dual Attention Prompter to automatically obtain prompt information through Spatial Attention and Channel Attention. We conduct comprehensive experiments on four public abdominal tumor segmentation datasets. The results indicate that our DEAP-3DSAM achieves state-of-the-art performance in 3D image segmentation, outperforming or matching existing manual prompt methods. Furthermore, both quantitative and qualitative ablation studies confirm the effectiveness of our proposed modules.

Citing comprehensively and appropriately has become a challenging task with the explosive growth of scientific publications. Current citation recommendation systems aim to recommend a list of scientific papers for a given text context or a draft paper. However, none of the existing work focuses on already included citations of full papers, which are imperfect and still have much room for improvement. In the scenario of peer reviewing, it is a common phenomenon that submissions are identified as missing vital citations by reviewers. This may lead to a negative impact on the credibility and validity of the research presented. To help improve citations of full papers, we first define a novel task of Recommending Missed Citations Identified by Reviewers (RMC) and construct a corresponding expert-labeled dataset called CitationR. We conduct an extensive evaluation of several state-of-the-art methods on CitationR. Furthermore, we propose a new framework RMCNet with an Attentive Reference Encoder module mining the relevance between papers, already-made citations, and missed citations. Empirical results prove that RMC is challenging, with the proposed architecture outperforming previous methods in all metrics. We release our dataset and benchmark models to motivate future research on this challenging new task.

Pre-trained Large Language Models (LLMs) have demonstrated prominent generalization to various linguistic tasks. However, due to the inherent modality and task discrepancy, parameter-efficient transfer learning for adapting LLMs to vision-language (VL) tasks remains challenging, which may struggle with excessive extra computation and data expenditure for VL pre-training and disconnection between multi-modal representations. This paper concentrates on the parameter-efficient adaptation of LLMs to VL tasks without inflexible multi-modal alignment pre-training on additional image-text pairs. Inspired by Instruction Tuning and the nature of multi-modal representation learning, we propose Multi-modal Prompt Tuning for Language Models (MPT4LM). This method provides text-relevant visual prompts via a plug-and-play Cross-Attention module and integrates them with textual Learnable Instruction as multi-modal prompts into LLMs. We further assemble MPT4LM with the currently prevalent Adapter approach to alleviate the trainable parameter scale and facilitate the collaboration of multi-modal prompts. We evaluate MPT4LM upon two representative LLMs: LLAMA-2 and Flan-T5, over two VL tasks: Visual Question Answering (VQAv2.0, GQA) and Visual Entailment (SNLI-VE). Extensive experimental results reveal that MPT4LM achieves state-of-the-art performance among prompting methods with only fine-tuning about 0.65% of the parameters of backbones, indicating a better trade-off between computation and data overhead and model performance. Our code is available at: https://github.com/YzM1a0/MPT4LM.

Automatically condensing multiple topic-related scientific papers into a succinct and concise summary is referred to as Multi-Document Scientific Summarization (MDSS). Currently, while commonly used abstractive MDSS methods can generate flexible and coherent summaries, the difficulty in handling global information and the lack of guidance during decoding still make it challenging to generate better summaries. To alleviate these two shortcomings, this paper introduces summary candidates into MDSS, utilizing the global information of the document set and additional guidance from the summary candidates to guide the decoding process. Our insights are twofold: Firstly, summary candidates can provide instructive information from both positive and negative perspectives, and secondly, selecting higher-quality candidates from multiple options contributes to producing better summaries. Drawing on the insights, we propose a summary candidates fusion framework - Disentangling Instructive information from Ranked candidates (DIR) for MDSS. Specifically, DIR first uses a specialized pairwise comparison method towards multiple candidates to pick out those of higher quality. Then DIR disentangles the instructive information of summary candidates into positive and negative latent variables with Conditional Variational Autoencoder. These variables are further incorporated into the decoder to guide generation. We evaluate our approach with three different types of Transformer-based models and three different types of candidates, and consistently observe noticeable performance improvements according to automatic and human evaluation. More analyses further demonstrate the effectiveness of our model in handling global information and enhancing decoding controllability.

Knowledge Tracing (KT) is the task of modeling a learner’s knowledge state to predict future performance in e-learning systems based on past performance. Deep learning-based methods, such as recurrent neural networks, memory-augmented neural networks, and attention-based neural networks, have recently been used in KT. Such methods have demonstrated excellent performance in capturing the latent dependencies of a learner’s knowledge state on recent exercises. However, these methods have limitations when it comes to dealing with the so-called Skill Switching Phenomenon (SSP), i.e., when learners respond to exercises in an e-learning system, the latent skills in the exercises typically switch irregularly. SSP will deteriorate the performance of deep learning-based approaches for simulating the learner’s knowledge state during skill switching, particularly when the association between the switching skills and the previously learned skills is weak. To address this problem, we propose the Memory-augmented Attentive Network (MAN), which combines the advantages of memory-augmented neural networks and attention-based neural networks. Specifically, in MAN, memory-augmented neural networks are used to model learners’ longer term memory knowledge, while attention-based neural networks are used to model learners’ recent term knowledge. In addition, we design a context-aware attention mechanism that automatically weighs the tradeoff between these two types of knowledge. With extensive experiments on several e-learning datasets, we show that MAN effectively improve predictive accuracies of existing state-of-the-art DLKT methods.

Prompt tuning, like CoOp, has recently shown promising vision recognizing and transfer learning ability on various downstream tasks with the emergence of large pre-trained vision-language models like CLIP. However, we identify that existing uni-modal prompt tuning approaches may result in sub-optimal performance since this uni-modal design breaks the original alignment of textual and visual representations in the pre-trained model. Inspired by the nature of pre-trained vision-language models, we aim to achieve completeness in prompt tuning and propose a novel approach called Multi-modal Deep-symphysis Prompt Tuning, dubbed as MuDPT, which extends independent multi-modal prompt tuning by additionally learning a model-agnostic transformative network to allow deep hierarchical bi-directional prompt fusion. We evaluate the effectiveness of MuDPT on few-shot vision recognition and out-of-domain generalization tasks. Compared with the state-of-the-art methods, MuDPT achieves better recognition and generalization ability with an apparent margin thanks to synergistic alignment of textual and visual representations. Our code is available at: https://github.com/Mechrev0/MuDPT.

Multi-Document Scientific Summarization (MDSS) aims to produce coherent and concise summaries for clusters of topic-relevant scientific papers. This task requires precise understanding of paper content and accurate modeling of cross-paper relationships. Knowledge graphs convey compact and interpretable structured information for documents, which makes them ideal for content modeling and relationship modeling. In this paper, we present KGSum, an MDSS model centred on knowledge graphs during both the encoding and decoding process. Specifically, in the encoding process, two graph-based modules are proposed to incorporate knowledge graph information into paper encoding, while in the decoding process, we propose a two-stage decoder by first generating knowledge graph information of summary in the form of descriptive sentences, followed by generating the final summary. Empirical results show that the proposed architecture brings substantial improvements over baselines on the Multi-Xscience dataset.

Citing comprehensively and appropriately has become a challenging task with the rapid growth of academic publications. Citation recommendation helps alleviate the burden of finding relevant and appropriate citations by recommending a list of academic papers for a given text. Existing approaches tried to determine whether a paper should be cited by measuring its relevance or similarity to a given citation circumstance. Citation circumstance modeling should consider many factors, such as the content of the citing paper, citation context words, citation network, and so on. However, most models only focus on citation context words and learn representations of citing paper through citation networks, putting little attention on citing paper content. In this paper, we regard the title as a summative and informative sentence relative to paper content and propose a novel model based on Bidirectional Gated Recurrent Units (BiGRUs) and Attentions. Our approach uses sequential embedding of paper title words for paper semantic representation and models citation circumstances by integrating citing paper title and citation context. Moreover, we introduce a semantic weight parameter to distinguish the importance of academic papers and citation contexts. Experiments on the ACL anthology network dataset show that our approach outperforms other state-of-the-art models in recall, MAP, MRR, and nDCG criteria.

Entity annotation in Wikipedia (officially named wikilinks) greatly benefits human end-users. Human editors are required to select all mentions that are most helpful to human end-users and link each mention to a Wikipedia page. We aim to design an automatic system to generate Wikipedia-style entity annotation for any plain text. However, existing research either rely heavily on mention-entity map or are restricted to named entities only. Besides, they neglect to select the appropriate mentions as Wikipedia requires. As a result, they leave out some necessary annotation and introduce excessive distracting annotation. Existing benchmarks also skirt around the coverage and selection issues. We propose a new task called Mention Detection and Se-lection for entity annotation, along with a new benchmark, WikiC, to better reflect annotation quality. The task is coined centering mentions specific to each position in high-quality human-annotated examples. We also proposed a new framework, DrWiki, to fulfill the task. We adopt a deep pre-trained span selection model inferring directly from plain text via tokens' context embedding. It can cover all possible spans and avoid limiting to mention-entity maps. In addition, information of both inarguable mention-entity pairs, and mention repeat has been introduced as token-wise representation enhancement by FLAT attention and repeat embedding respectively. Empirical results on WikiC show that, compared with often adopted and state-of-the-art Entity Linking and Entity Recognition methods, our method achieves improvement to previous methods in overall performance. Additional experiments show that DrWiki gains improvement even with a low-coverage mention-entity map.

Current state-of-the-art approaches for image captioning mainly apply the encoder-decoder framework with attention mechanisms, most of which ignore interactions between different types of image features and perform attention operations only once per word. The mentioned problems limit the captioning model’s capability to capture sufficient information to generate high-quality captions. By contrast, humans often rethink to polish up descriptions by re-focusing on more correct and important information, which is hard to capture at first glance. In this paper, we introduce a novel topic-guided captioning model to imitate such a human’s rethinking process by modeling interactions between visual and hierarchical semantic features of topics. To the best of our knowledge, we are the first to effectively consider hierarchical semantic features as guidance to facilitate visual attention, achieving human-like rethinking for captioning. Extensive experiments on the MS COCO dataset show that our proposed model achieves superior performance over state-of-the-art methods.

Image captioning requires models to translate an image into natural language description. Recent captioning methods are mostly based on the popular encoder-decoder framework, which uses the encoder to encode an image into a set of vectors and a decoder to produce the caption word-by-word. However, when generating the next word, the decoder only looks back at the preceding words but fails to foresee the posterior ones that are supposed to be produced, lacking awareness of the global semantic information of the image. Moreover, the existing generated caption provides global semantic information of a given image but sometimes contains inconsistent content that needs to be fixed. This paper proposes a novel Temporal-Free Semantic-Guided attention mechanism (TFSG) to utilize the raw caption pre-generated by a primary decoder as the extra input to provide global semantic guidance during generation, deepening visual understanding by balancing the semantic and visual information. Specifically, we design two attention-based structures for TFSG to selectively take the essence from the semantic and visual information to produce refined descriptions. We apply TFSG to two popular captioning models and conduct extensive experiments on the public MS COCO benchmark. Results show that our proposed model achieves superior performance over many state-of-the-art models.

Depression is a common mental disease, and it is committed to promote the research of depression assessment based on physiological signals. By collecting the resting state EEG data of depressive disorder, we collected the resting state EEG data of 14 patients with depression and 17 normal people. Through the analysis, the number of troughs of each person's data was statistically analyzed, combined with the convolution neural network model. The accuracy rate of some data is 94.88% by the statistical trough number method, and the accuracy rate of the remaining part of the test data set is 85.0% through the convolution neural network model, and the final fusion accuracy rate is 86.88%. The experimental results show that the combination of statistical trough number and convolution neural network can distinguish depression patients better in resting state EEG data.

During the past five years, tons of economic 360 VR cameras (e.g., Ricoh Theta, Samsumg Gear360, LG 360, Insta 360) are sold in the market. While 360 VR videos become ubiquitous very soon, 360 VR video standardization is still under discussion in the digital industry, and more concrete efforts are desired to accelerate its standardization and applications. Though ERP has been widely used for projection and packing layout while encoding 360 VR videos, it has severe projection distortion near poles. In this paper, we introduce a new format for encoding and storing 360 VR videos using hybrid cylindrical projection after thoroughly analyzing the problems with ERP. We show that our new hybrid format can minimize stretching distortion and generate well balanced pixel distribution in the resulting projection.

An important aspect of communication in Twitter (and other Social Network is message propagation -- people creating posts for others to share. Although there has been work on modelling how tweets in Twitter are propagated (retweeted), an untackled problem has been who will retweet a message. Here we consider the task of finding who will retweet a message posted on Twitter. Within a learning to-rank framework, we explore a wide range of features, such as retweet history, followers status, followers active time and followers interests. We find that followers who retweeted or mentioned the author's tweets frequently before and have common interests are more likely to be retweeters.

2011-2012 > Academic research: refereed > Refereed conference paper

The key issue of person name disambiguation is to discover different namesakes in massive web documents rather than simply cluster documents by using textual features. In this paper, we describe a novel person name disambiguation method based on social networks to effectively identify namesakes. The social network snippets in each document are extracted. Then, the namesakes are identified via splicing the social networks of each namesake by using the snippets as a bipartite graph. Experimental results show that our method achieves better result than the top performance of WePS-2 in identifying different namesakes.