Shaoning Sun, Mingzhu Cai, Huang He, Bingjin Chen, Siqi Bao, Yujiu Yang, Hua Wu, Haifeng Wang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

In the processing of high-resolution remote sensing images, multiscale feature fusion techniques are commonly employed to construct change detection models, aiming to capture the details and characteristics of target objects at different scales. However, current change detection methods often neglect the subtle features of small-scale targets and edge information during feature fusion. To address this issue, this article proposes a deep feature fusion network (DFFNet) for full-scale change detection in remote sensing images. DFFNet enhances the accuracy of boundary information in change regions and enables the extraction of full-scale change features. By integrating phased difference extraction techniques with an enhanced attention mechanism, a dual temporal difference enhancement module (DT-DEM) is designed to comprehensively extract change detection information across scales. In addition, combining an inverted pyramid technique with an attention mechanism, a screening function inverted pyramid network (S-FIPN) is proposed, which efficiently extracts and fuses full-scale features while significantly improving the extraction of subtle changes and weak edge features. Comparison experiments with ten state-of-the-art (SOTA) algorithms on four publicly available datasets, namely, LEVIR-CD, WHU-CD, NJDS, and MSRS-CD, show that DFFNet achieves the best$F1$values on all datasets, which are 92.50%, 90.55%, 81.04%, and 77.44%, respectively. These results underscore the superiority and effectiveness of DFFNet in full-scale feature extraction and edge information retrieval.

The Track-1 of DSTC9 aims to effectively answer user requests or questions during task-oriented dialogues, which are out of the scope of APIs/DB. By leveraging external knowledge resources, relevant information can be retrieved and encoded into the response generation for these out-of-API-coverage queries. In this work, we have explored several advanced techniques to enhance the utilization of external knowledge and boost the quality of response generation, includingschema guided knowledge decision,negatives enhanced knowledge selection, andknowledge grounded response generation. To evaluate the performance of our proposed method, comprehensive experiments have been carried out on the publicly available dataset. Our approach was ranked as the best in human evaluation of DSTC9 Track-1.

In recent years, Internet memes have been widely used in online chatting. Compared with text-based communication, conversations become more expressive and attractive when Internet memes are incorporated. This article presents our solutions for the Meme incorporated Open-domain Dialogue (MOD) Challenge of DSTC10, where three tasks are involved: text response modeling, meme retrieval, and meme emotion classification. Firstly, we leverage a large-scale pre-trained dialogue model for coherent and informative response generation. Secondly, based on interaction-based text-matching, our approach can retrieve appropriate memes with good generalization ability. Thirdly, we propose to model the emotion flow (EF) in conversations and introduce an auxiliary task of emotion description prediction (EDP) to boost the performance of meme emotion classification. Experimental results on the MOD dataset demonstrate that our methods can incorporate Internet memes into dialogue systems effectively.

Many open-domain dialogue models pretrained with social media comments can generate coherent replies but have difficulties producing engaging responses.This phenomenon might mainly result from the deficiency of annotated human-human conversations and the misalignment with human preference.In this paper, we propose a novel and efficient framework Diamante to boost the open-domain chatbot, where two kinds of human feedback (including explicit demonstration and implicit preference) are collected and leveraged.By asking annotators to select or amend the modelgenerated candidate responses, Diamante efficiently collects the human demonstrated responses and constructs a Chinese chit-chat dataset.To enhance the alignment with human preference, Diamante leverages the implicit preference in the data collection process and introduces the generation-evaluation joint training.Comprehensive experiments indicate that the Diamante dataset and joint training paradigm can significantly boost the performance of pre-trained dialogue models.The overall engagingness of the previous state-ofthe-art model has been improved remarkably by 50% in Chinese open-domain conversations.

In this paper, we propose an unsupervised query enhanced approach for knowledgeintensive conversations, namely QKConv.There are three modules in QKConv: a query generator, an off-the-shelf knowledge selector, and a response generator.QKConv is optimized through joint training, which produces the response by exploring multiple candidate queries and leveraging corresponding selected knowledge.The joint training solely relies on the dialogue context and target response, getting exempt from extra query annotations or knowledge provenances.To evaluate the effectiveness of the proposed QKConv, we conduct experiments on three representative knowledgeintensive conversation datasets: conversational question-answering, task-oriented dialogue, and knowledge-grounded conversation.Experimental results reveal that QKConv performs better than all unsupervised methods across three datasets and achieves competitive performance compared to supervised methods.

Existing pipelined task-oriented dialogue systems usually have difficulties adapting to unseen domains, whereas end-to-end systems are plagued by large-scale knowledge bases in practice.In this paper, we introduce a novel querydriven task-oriented dialogue system, namely Q-TOD.The essential information from the dialogue context is extracted into a query, which is further employed to retrieve relevant knowledge records for response generation.Firstly, as the query is in the form of natural language and not confined to the schema of the knowledge base, the issue of domain adaption is alleviated remarkably in Q-TOD.Secondly, as the query enables the decoupling of knowledge retrieval from the generation, Q-TOD gets rid of the issue of knowledge base scalability.To evaluate the effectiveness of the proposed Q-TOD, we collect query annotations for three publicly available task-oriented dialogue datasets.Comprehensive experiments verify that Q-TOD outperforms strong baselines and establishes a new state-of-the-art performance on these datasets.

Pre-training models have been proved effective for a wide range of natural language processing tasks.Inspired by this, we propose a novel dialogue generation pre-training framework to support various kinds of conversations, including chit-chat, knowledge grounded dialogues, and conversational question answering.In this framework, we adopt flexible attention mechanisms to fully leverage the bi-directional context and the uni-directional characteristic of language generation.We also introduce discrete latent variables to tackle the inherent one-to-many mapping problem in response generation.Two reciprocal tasks of response generation and latent act recognition are designed and carried out simultaneously within a shared network.Comprehensive experiments on three publicly available datasets verify the effectiveness and superiority of the proposed framework.

High efficient video coding (HEVC) is a new video coding compression standard. HEVC adopts context-based adaptive binary arithmetic coding (CABAC) as the entropy coding scheme. In this paper, the overall architecture and efficiency of the main frequency are improved by the optimization of the input and output modules and the module optimization of the arithmetic coding CABAC hardware structure. In terms of input module optimization, four-level buffer input and residual coefficient transmission optimization are adopted; in terms of arithmetic coding module optimization, context model index pre-reading, pre-normalization look-up table and in-line serial stream output design are adopted so as to improve the overall efficiency of the architecture and the main frequency, reduce resource consumption, and achieve a high-frequency hardware architecture of the efficient coding pipeline. The combined results show that the pipeline can operate at 370MHz with 43.49K gates aiming at 90nm process. The processing rate and throughput can support real-time encoding of 1080P video under the general test conditions of the HEVC standard of 30 frames per second.

In this paper, a novel Generation-Evaluation framework is developed for multi-turn conversations with the objective of letting both participants know more about each other.For the sake of rational knowledge utilization and coherent conversation flow, a dialogue strategy which controls knowledge selection is instantiated and continuously adapted via reinforcement learning.Under the deployed strategy, knowledge grounded conversations are conducted with two dialogue agents.The generated dialogues are comprehensively evaluated on aspects like informativeness and coherence, which are aligned with our objective and human instinct.These assessments are integrated as a compound reward to guide the evolution of dialogue strategy via policy gradient.Comprehensive experiments have been carried out on the publicly available dataset, demonstrating that the proposed method outperforms the other state-of-the-art approaches significantly.

The split-delivery vehicle routing problem (SDVRP) is a natural extension of the classical vehicle routing problem (VRP) that allows the same customer to be served by more than one vehicle. This problem is a very challenging combinatorial optimization problem and has attracted much academic attention. To solve it, most of the literature articles adopted heuristic approaches in which the solution is represented by a set of delivery patterns, and the search operators were derived from the traditional VRP operators. Differently, our approach employs the combination of a set of routes and a forest to represent the solution. Several forest-based operators are accordingly introduced. We integrate the new operators into a simple tabu search framework and then demonstrate the efficiency of our approach by conducting experiments on existing benchmark instances.