Zero-shot visual question answering (VQA) requires models to reason over unseen image–question pairs without task-specific supervision, where performance is often undermined by incomplete visual grounding and unstable reasoning. Recent approaches attempt to mitigate this by using large language models to generate external knowledge conditioned on captions and questions. However, these methods typically operate in an open-loop manner, lacking mechanisms to regulate competing interpretations or correct misaligned knowledge. Inspired by the hippocampal pattern completion mechanism, which supports inference from partial observations through memory reactivation and competitive stabilization, we propose ARK-PC for zero-shot VQA, which formulates reasoning as a two-stage completion process. It first expands fragmented multimodal cues into multiple structured candidate hypotheses, then adaptively reinforces coherent candidates while suppressing inconsistent ones through iterative feedback. By coupling knowledge generation with competitive refinement, ARK-PC transforms open-loop inference into a closed-loop stabilization process, enabling robust reasoning under uncertainty without external supervision. Experiments on OK-VQA and A-OKVQA demonstrate consistent state-of-the-art zero-shot performance and strong generalization across diverse backbones, indicating that the improvements stem from the proposed framework rather than model scale.

This paper introduces HQT-TI, a novel indexing method designed to improve the efficiency of spatial keyword queries. HQT-TI consists of two main components: a Hilbert QuadTree (HQT) based spatial index and a Trie-Inverted index (TI) combined textual index. HQT integrates the Hilbert curve with a Quadtree, establishing a direct relationship between the two. TI combines a trie and inverted index to minimize the intersection cost associated with long lists, thus improving the speed of keyword queries. The HQT based Spatial Query algorithm (HQT-SQ) reduces overlap checks and limits irrelevant object retrieval by employing query drill-down and depth first search with limited breadth expansion in spatial queries. Meanwhile, the Segment List Intersection based Keyword Query algorithm (SLI-KQ), built on TI, efficiently handles segment list intersections for keyword queries. The combination of HQT-SQ and SLI-KQ results in HS-SK, a highly efficient spatial keyword query algorithm. Extensive experimental results demonstrate that HS-SKQ outperforms SFC-Quad by up to two orders of magnitude, achieving up to a 5.46× speedup over the best existing competitors, making it a promising solution for large-scale spatial keyword query processing.

Document-level relation extraction (DocRE) is the process of identifying and extracting relations between entities that span multiple sentences within a document. Due to its realistic settings, DocRE has garnered increasing research attention in recent years. Previous research has mostly focused on developing sophisticated encoding models to better capture the intricate patterns between entity pairs. While these advancements are undoubtedly crucial, an even more foundational challenge lies in the data itself. The complexity inherent in DocRE makes the labeling process prone to errors, compounded by the extreme sparsity of positive relation samples, which is driven by both the limited availability of positive instances and the broad diversity of positive relation types. These factors can lead to biased optimization processes, further complicating the task of accurate relation extraction. Recognizing these challenges, we have developed a robust framework called COMM to better solve DocRE. COMM operates by initially employing an instance-aware reasoning method to dynamically capture pertinent information of entity pairs within the document and extract relational features. Following this, COMM takes into account the distribution of relations and the difficulty of samples to dynamically adjust the margins between prediction logits and the decision threshold, a process we call Concentrated Margin Maximization. In this way, COMM not only enhances the extraction of relevant relational features but also boosts DocRE performance by addressing the specific challenges posed by the data. Extensive experiments and analysis demonstrate the versatility and effectiveness of COMM, especially its robustness when trained on low-quality data (achieves >10% performance gains).

Tengyu Pan, Zhichao Duan, Zhenyu Li, Bowen Dong, Ning Liu, Xiuxing Li, Jianyong Wang. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

fMRI (functional Magnetic Resonance Imaging) visual decoding involves decoding the original image from brain signals elicited by visual stimuli. This often relies on manually labeled ROIs (Regions of Interest) to select brain voxels. However, these ROIs can contain redundant information and noise, reducing decoding performance. Additionally, the lack of automated ROI labeling methods hinders the practical application of fMRI visual decoding technology, especially for new subjects. This work presents TROI (Trainable Region of Interest), a novel two-stage, data-driven ROI labeling method for cross-subject fMRI decoding tasks, particularly when subject samples are limited. TROI leverages labeled ROIs in the dataset to pretrain an image decoding backbone on a cross-subject dataset, enabling efficient optimization of the input layer for new subjects without retraining the entire model from scratch. In the first stage, we introduce a voxel selection method that combines sparse mask training and low-pass filtering to quickly generate the voxel mask and determine input layer dimensions. In the second stage, we apply a learning rate rewinding strategy to fine-tune the input layer for downstream tasks. Experimental results on the same small sample dataset as the baseline method for brain visual retrieval and reconstruction tasks show that our voxel selection method surpasses the state-of-the-art method MindEye2 with an annotated ROI mask.

Set Similarity Query (SSQ) is widely applied in various fields. The existing B+-tree-based SSQ approaches fail to fully exploit length filtering and require calculating similarity bounds in a node-wise manner, leading to low efficiency. To address these issues, we propose LeB, a novel length-enhanced B+-tree index, whose keys integrate set lengths and bucket mapping, enabling the direct pruning of sets that do not meet the length requirements. Building upon LeB, we present an efficient algorithm, LeBQ, which leverages length filtering and symmetric difference allocation to determine the key bounds for a query, enabling the key bounds computation only once for each query$Q$and avoiding costly similarity bounds computation in a node-wise manner. Efficient key filtering strategies are proposed to prune sets that cannot be similar, significantly reducing the number of candidates. Based on LeBQ, LeBQ+ further reduces the number of candidates by introducing length-independent key bounds. Experimental results on four real datasets demonstrate that LeBQ+ has a higher node access efficiency and accesses only 3.08% to 27.47% nodes compared to the existing B+-tree-based SSQ algorithm. LeBQ+is up to 99.8 × faster than the state-of-the-art algorithms.

Motion planning in high-dimensional continuous spaces remains challenging due to complex environments and computational constraints. Although learning-based planners, especially graph neural network (GNN)-based, have significantly improved planning performance, they still struggle with inaccurate graph construction and limited structural reasoning, constraining search efficiency and path quality. The human brain exhibits efficient planning through a two-stage Perception-Decision model. First, egocentric spatial representations from visual and proprioceptive input are constructed, and then semantic–episodic synergy is leveraged to support decision-making in uncertainty scenarios. Inspired by this process, we propose NeuroMP, a brain-inspired planning framework that learns to plan like the human brain. NeuroMP integrates a Perceptive Segment Selector inspired by visuospatial perception to construct safer graphs, and a Global Alignment Heuristic guide search in weakly connected graphs by modeling semantic-episodic synergistic decision-making. Experimental results demonstrate that NeuroMP significantly outperforms existing planning methods in efficiency and quality while maintaining a high success rate.

Open-world reinforcement learning challenges agents to develop intelligent behavior in vast exploration spaces. Recent approaches like LS-Imagine have advanced the field by extending imagination horizons through jumpy state transitions, yet remain limited by fixed exploration mechanisms and static jump thresholds that cannot adapt across changing task phases, resulting in inefficient exploration and lower completion rates. Humans demonstrate remarkable capabilities in open-world decision-making through a chain-like process of task decomposition, selective memory utilization, and adaptive uncertainty regulation. Inspired by human decision-making processes, we present Cognitive Predictive Processing (CPP), a novel framework that integrates three neurologically-inspired systems: a phase-adaptive cognitive controller that dynamically decomposes tasks into exploration, approach, and completion phases with adaptive parameters; a dual-memory integration system implementing dual-modal memory that balances immediate context with selective long-term storage; and an uncertainty-modulated prediction regulator that continuously updates environmental predictions to modulate exploration behavior. Comprehensive experiments in MineDojo demonstrate that these human-inspired decision-making strategies enhance performance over recent techniques, with success rates improving by an average of 4.6\% across resource collection tasks while reducing task completion steps by an average of 7.1\%. Our approach bridges cognitive neuroscience and reinforcement learning, excelling in complex scenarios that require sustained exploration and strategic adaptation while demonstrating how neural-inspired models can solve key challenges in open-world AI systems.

Research on emotion recognition in conversations emphasises the importance of complex relationships between conversational context and multimodality. Graph-based methods, particularly hypergraph-based methods have shown promise in capturing these relationships. However, challenges persist in avoiding redundant context while capturing essential information for optimal context embeddings and fully leveraging cross-modal complementarities for sufficient fusion. In contrast, the human brain flexibly retrieves relevant memories and integrates multi-modal data for accurate recognition. Based on this superiority, we propose BrainyHGNN, a brain-inspired hypergraph neural network. It integrates a Dynamic Memory Selector for contextual hyperedges, mimicking selective memory retrieval mechanisms for adaptive and modality-specific context retrieval. HierSensNet is designed for multi-modal hyperedges, mirroring hierarchical cross-modal interaction mechanisms to ensure effective multimodal fusion. Experimental results on two benchmark datasets validate the superior performance of BrainyHGNN, confirming the effectiveness of its innovative approach. This work highlights the potential of brain-inspired methods to advance flexible context retrieval and sufficient multimodal fusion, presenting a promising direction for future research in this domain.

Efficient and reliable motion planning is essential for robots in transportation systems. Learning-based motion planners, especially those leveraging graph neural networks (GNNs), have emerged as promising approaches to accelerate motion planning. However, current GNN-based planners suffer from inevitable performance bottlenecks due to their insufficient exploitation of environmental information and the intrinsic connections within graph structures. In contrast, the human brain exhibits innate strengths in decision-making and reasoning, enabling complex inferences from sparse observations and rapid integration of new information to control behavior. Neuroscience evidence reveals the human brain translates decision-making problems into graph structures and sensory observations, leveraging spatial relational memory for sensory inference. To address these issues, this paper proposes a brain-inspired GNN-based motion planner, BrainyMP, which innovatively draws on the brain’s spatial relational memory mechanisms. Specifically, a selective sampling strategy is proposed to reduce unnecessary exploration during the construction of the random geometric graph (RGG). Additionally, a Brainy Edge Selector is designed to filter inappropriate edges, enhancing planning quality. Furthermore, the Memory-aware Predictor is proposed to improve graph pattern learning capabilities and planning efficiency by integrating subgraph structures. Extensive experimental results demonstrate that the proposed method significantly improves planning quality and efficiency in maze and robotic-arm manipulation tasks while maintaining high success rates. These results indicate that emulating human brain mechanisms holds promise for improving robotic performance.

Knowledge base question answering (KBQA) is a critical yet challenging task due to the vast number of entities within knowledge bases and the diversity of natural language questions posed by users. Unfortunately, the performance of most KBQA models tends to decline significantly in real-world scenarios where high-quality annotated data is insufficient. To mitigate the burden associated with manual annotation, we introduce FlexKBQA by utilizing Large Language Models (LLMs) as program translators for addressing the challenges inherent in the few-shot KBQA task. Specifically, FlexKBQA leverages automated algorithms to sample diverse programs, such as SPARQL queries, from the knowledge base, which are subsequently converted into natural language questions via LLMs. This synthetic dataset facilitates training a specialized lightweight model for the KB. Additionally, to reduce the barriers of distribution shift between synthetic data and real user questions, FlexKBQA introduces an executionguided self-training method to iterative leverage unlabeled user questions. Furthermore, we explore harnessing the inherent reasoning capability of LLMs to enhance the entire framework. Consequently, FlexKBQA delivers substantial flexibility, encompassing data annotation, deployment, and being domain agnostic. Through extensive experiments on GrailQA, WebQSP, and KQA Pro, we observe that under the few-shot even the more challenging zero-shot scenarios, FlexKBQA achieves impressive results with a few annotations, surpassing all previous baselines and even approaching the performance of supervised models, achieving a remarkable 93% performance relative to the fully-supervised models. We posit that FlexKBQA represents a significant advancement towards exploring better integration of large and lightweight models. Code is available at https://github.com/leezythu/FlexKBQA.

Task-oriented dialog systems have witnessed substantial progress due to conversational pre-training techniques. Yet, two significant challenges persist. First, most systems primarily utilize the latest turn's state label for the generator. This practice overlooks the comprehensive value of state labels in boosting the model's understanding for future generations. Second, an overreliance on generated policy often leads to error accumulation, resulting in suboptimal responses when adhering to incorrect actions. To combat these challenges, we propose turn-level multi-task objectives for the encoder. With the guidance of essential information from labeled intermediate states, we establish a more robust representation for both understanding and generation. For the decoder, we introduce an action tree-based scheduled sampling technique. Specifically, we model the hierarchical policy as trees and utilize the similarity between trees to sample negative policy based on scheduled sampling, hoping the model to generate invariant responses under perturbations. This method simulates potential pitfalls by sampling similar negative policy, bridging the gap between task-oriented dialog training and inference. Among methods without continual pre-training, our approach achieved state-of-the-art (SOTA) performance on the MultiWOZ dataset series and was also competitive with pre-trained SOTA methods.

The ever-growing biomedical publications magnify the challenge of extracting structured data from unstructured texts.This task involves two components: biomedical entity identification (Named Entity Recognition, NER) and their interrelation determination (Relation Extraction, RE).However, existing methods often neglect unique features of the biomedical literature, such as ambiguous entities, nested proper nouns, and overlapping relation triplets, and underutilize prior knowledge, leading to an intolerable performance decline in the biomedical domain, especially with limited annotated training data.In this paper, we propose the Biomedical Relation-First eXtraction (Bio-RFX) model by leveraging sentence-level relation classification before entity extraction to tackle entity ambiguity.Moreover, we exploit structural constraints between entities and relations to guide the model's hypothesis space, enhancing extraction performance across different training scenarios.Comprehensive experimental results on biomedical datasets show that Bio-RFX achieves significant improvements on both NER and RE tasks.Even under the lowresource training scenarios, it outperforms all baselines in NER and has highly competitive performance compared to the state-of-the-art fine-tuned baselines in RE 1 .

Large language models (LLMs) offer impressive performance across diverse fields, but their increasing complexity raises both design costs and the need for specialized expertise. These challenges are intensified for Neural Architecture Search (NAS) methods reliant on weight-sharing techniques. This paper introduces GNAS, a new NAS method that boosts the search process with the aid of LLMs for efficient model discovery. With insights from existing architectures, GNAS swiftly identifies superior models that can adapt to changing resource constraints. We provide a mathematical framework to facilitate the transfer of knowledge across different model sizes, thereby improving search efficiency. Our experiments conducted on ImageNet, NAS-Bench-Macro, and ChannelBench-Macro confirm the effectiveness of GNAS across both CNN and Transformer architectures.

A major challenge in motor imagery Brain-Computer Interfaces (MI-BCIs) arises from domain shift due to large individual differences. Currently, most cross-subject MI-BCI decoding methods rely on transfer learning to extract subject-shared features or align data distributions. However, these methods typically require all unlabeled data from the target subjects or labeled calibration data, which is unavailable in practical applications. To address this, we propose a brain lateralization-guided subject adaptive network, BLSAN, to enhance model generalization through local-global adversarial training. Specifically, two separate adversarial networks for left and right hemispheres are designed to reduce local differences, and features extracted from both hemispheres are combined for global adversarial training. Additionally, we design a confidence-based pseudo label generation method to enhance model discriminability. We validate the effectiveness of our approach on two public MI datasets, BCI Competition IV 2a and 2b, only with some unlabeled calibration data, which improves the practicality of MI-BCIs.

Although advances in deep learning technologies have greatly facilitated the brain intention decoding from electroencephalogram (EEG) in motor imagery brain–computer interfaces (MI-BCIs), significant individual differences hinder the practical cross-subject MI-BCI applications. Unlike other existing domain adversarial transfer networks that focus on designing different discriminators to reduce individual differences, inspired by the motor lateralization phenomenon, we innovatively utilize transformer and the spatiotemporal pattern differences of EEG as prior knowledge to enhance the feature discriminability in our brain decoding adversarial network. In addition, to address adversarial network decision boundaries bias toward the source domain, we propose a data augmentation method, EEGMix to rapidly mix and enrich the target domain data. With an adaptive adversarial factor, our decoding model reduces the differences in marginal and conditional distribution simultaneously. Three public MI datasets, 2a, 2b, and OpenBMI verified our model's effectiveness. The accuracy achieved 77.49%, 85.19%, and 79.37%, superior to other state-of-the-art algorithms.

Ever-growing electronic medical corpora provide unprecedented opportunities for researchers to analyze patient conditions and drug effects. Meanwhile, severe challenges emerged in the large-scale electronic medical records process phase. Primarily, emerging words for medical terms, including informal descriptions, are difficult to recognize. Moreover, although deep models can help in entity extraction on medical texts, they require large-scale labels, which are time-intensive to obtain and not always available in the medical domain. However, when encountering a situation where massive unseen concepts appear or labeled data is insufficient, the performance of existing algorithms will suffer an intolerable decline. In this article, we propose a balanced and deep active learning framework for Medical Named Entity Recognition (MedNER) to alleviate the above problems. Specifically, to describe our selection strategy precisely, we first define the uncertainty of a medical sentence as a labeling loss predicted by a loss-prediction module and define diversity as the least text distance between pairs of sentences in a sample batch computed based on word-morpheme embeddings. Furthermore, aiming to make a trade-off between uncertainty and diversity, we formulate a Distinct-K optimization problem to maximize the slightest uncertainty and diversity of chosen sentences. Finally, we propose a threshold-based approximation selection algorithm, Distinct-K Filter , which selects the most beneficial training samples by balancing diversity and uncertainty. Extensive experimental results on real datasets demonstrate that MedNER significantly outperforms existing approaches.

Structured tabular data is a fundamental data type in numerous fields, and the capacity to reason over tables is crucial for answering questions and validating hypotheses. However, constructing labeled data for complex reasoning tasks is labor-intensive, and the quantity of annotated data remains insufficient to support the intricate demands of real-world applications. To address the insufficient annotation challenge, we present a self-training framework for unsupervised complex tabular reasoning (UCTR-ST) by generating diverse synthetic data with complex logic. Specifically, UCTR-ST incorporates several essential techniques: we aggregate diverse programs and execute them on tables based on a “Program-Management” component, and we bridge the gap between programs and text with a powerful “Program-Transformation” module that generates natural language sentences with complex logic. Furthermore, we optimize the procedure using “Table-Text Manipulator” to handle joint table-text reasoning scenarios. The entire framework utilizes self-training techniques to leverage the unlabeled training data, which results in significant performance improvements when tested on real-world data. Experimental results demonstrate that UCTR-ST achieves above 90% of the supervised model performance on different tasks and domains, reducing the dependence on manual annotation. Additionally, our approach can serve as a data augmentation technique, significantly boosting the performance of supervised models in low-resourced domains.

Structured tabular data exist across nearly all fields. Reasoning task over these data aims to answer questions or determine the truthiness of hypothesis sentences by understanding the semantic meaning of a table. While previous works have devoted significant efforts to the tabular reasoning task, they always assume there are sufficient labeled data. However, constructing reasoning samples over tables (and related text) is labor-intensive, especially when the reasoning process is complex. When labeled data is insufficient, the performance of models will suffer an unendurable decline. In this paper, we propose a unified framework for unsupervised complex tabular reasoning (UCTR), which generates sufficient and diverse synthetic data with complex logic for tabular reasoning tasks, assuming no human-annotated data at all. Specifically, we first utilize a random sampling strategy to collect diverse programs of different types and execute them on tables based on a "Program-Executor" module. To bridge the gap between the programs and natural language sentences, we design a powerful "NL-Generator" module to generate natural language sentences with complex logic from these programs. Since a table often occurs with its surrounding texts, we further propose novel "Table-to-Text" and "Text-to-Table" operators to handle joint table-text reasoning scenarios. This way, we can adequately exploit the unlabeled table resources to obtain a well-performed reasoning model under an unsupervised setting. Our experiments cover different tasks (question answering and fact verification) and different domains (general and specific), showing that our unsupervised methods can achieve at most 93% performance compared to supervised models. The impressive performance demonstrates that UCTR can significantly reduce the dependence on manual annotation. Moreover, we also find that it can substantially boost the supervised performance in low-resourced domains as a data augmentation technique.

Entity linking aims to link ambiguous mentions to their corresponding entities in a knowledge base, which is significant and fundamental for various downstream applications, e.g., knowledge base completion, question answering, and information extraction. While great efforts have been devoted to this task, most of these studies follow the assumption that large-scale labeled data is available. However, when the labeled data is insufficient for specific domains due to labor-intensive annotation work, the performance of existing algorithms will suffer an intolerable decline. In this paper, we endeavor to solve the problem of few-shot entity linking, which only requires a minimal amount of in-domain labeled data and is more practical in real situations. Specifically, we firstly propose a novel weak supervision strategy to generate non-trivial synthetic entity-mention pairs based on mention rewriting. Since the quality of the synthetic data has a critical impact on effective model training, we further design a meta-learning mechanism to assign different weights to each synthetic entity-mention pair automatically. Through this way, we can profoundly exploit rich and precious semantic information to derive a well-trained entity linking model under the few-shot setting. The experiments on real-world datasets show that the proposed method can extensively improve the state-of-the-art few-shot entity linking model and achieve impressive performance when only a small amount of labeled data is available. Moreover, we also demonstrate the outstanding ability of the model's transferability. Our code and models will be open-sourced.

Similarity search is an essential operation in many applications. Given a collection of set records and a query, the exact set similarity search aims at finding all the records that are similar to the query from the collection. Existing methods adopt a filter-and-verify framework, which make use of inverted indexes. However, as the complexity of verification is rather low for setbased similarity metrics, they always fail to make a good tradeoff between filter power and filter cost. In this paper, we proposed an efficient framework for exact set similarity search based on tree index structure. We defined a hash-based ordering to effectively import data into the index structure and then make optimizations to reduce the filter cost. To further improve the filter power, we proposed a dynamic algorithm to partition the dataset into several parts and propose a multiple-index framework. Experimental results on real-world datasets show that our method significantly outperform the state-of-the-art algorithms.