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.

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.

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.

Spam and wildly varying documents make searching in Twitter challenging. Most Twitter search systems generally treat a Tweet as a plain text when modeling relevance. However, a series of conventions allows users to Tweet in structural ways using a combination of different blocks of texts. These blocks include plain texts, hashtags, links, mentions, etc. Each block encodes a variety of communicative intent and the sequence of these blocks captures changing discourse. Previous work shows that exploiting the structural information can improve the structured documents (e.g., web pages) retrieval. In this study we utilize the structure of Tweets, induced by these blocks, for Twitter retrieval and Twitter opinion retrieval. For Twitter retrieval, a set of features, derived from the blocks of text and their combinations, is used into a learning‐to‐rank scenario. We show that structuring Tweets can achieve state‐of‐the‐art performance. Our approach does not rely on social media features, but when we do add this additional information, performance improves significantly. For Twitter opinion retrieval, we explore the question of whether structural information derived from the body of Tweets and opinionatedness ratings of Tweets can improve performance. Experimental results show that retrieval using a novel unsupervised opinionatedness feature based on structuring Tweets achieves comparable performance with a supervised method using manually tagged Tweets. Topic‐related specific structured Tweet sets are shown to help with query‐dependent opinion retrieval.

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.

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.

Through the investigation of email document structure, this paper proposes a multi-field learning (MFL) framework, which breaks the multi-field document Text Classification (TC) problem into several sub-document TC problems, and makes the final category prediction by weighted linear combination of several sub-document TC results. Many previous statistical TC algorithms can be easily rebuilt within the MFL framework via turning binary result to spamminess score, which is a real number and reflects the likelihood that the classified email is spam. The experimental results in the TREC spam track show that the performances of many TC algorithms can be improved within the MFL framework.

Creating domain-specific ontologies is one of the main bottlenecks in the development of the semantic Web. Learning an ontology from linguistic resources is helpful to reduce the costs of ontology creation. In this paper, we describe a method to extract the most related concepts from HowNet, a Chinese-English bilingual knowledge dictionary, in order to create a customized ontology for a particular domain. We introduce a new method to measure relatedness (rather than similarity between concepts), which overcomes some of the traditional problems associated with similar concepts being far apart in the hierarchy. Experiments show encouraging results.