Jingxuan Wei, Xingyue Wang, Yanghaoyu Liao, Jie Dong, Yuchen Liu, Caijun Jia, Bihui Yu, Junnan Zhu. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Linzhuang Sun, Hao Liang, Jingxuan Wei, Bihui Yu, Tianpeng Li, Fan Yang, Zenan Zhou, Wentao Zhang. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

We introduce the task of text-to-diagram generation, which focuses on creating structured visual representations directly from textual descriptions. Existing approaches in text-to-image and text-to-code generation lack the logical organization and flexibility needed to produce accurate, editable diagrams, often resulting in outputs that are either unstructured or difficult to modify. To address this gap, we introduce DiagramGenBenchmark, a comprehensive evaluation framework encompassing eight distinct diagram categories, including flowcharts, model architecture diagrams, and mind maps. Additionally, we present DiagramAgent, an innovative framework with four core modules—Plan Agent, Code Agent, Check Agent, and Diagram-to-Code Agent—designed to facilitate both the generation and refinement of complex diagrams. Our extensive experiments, which combine objective metrics with human evaluations, demonstrate that DiagramAgent significantly outperforms existing baseline models in terms of accuracy, structural coherence, and modifiability. This work not only establishes a foundational benchmark for the text-to-diagram generation task but also introduces a powerful toolset to advance research and applications in this emerging area.

Chart question answering (CQA) has become a critical multimodal task for evaluating the reasoning capabilities of vision-language models.While early approaches have shown promising performance by focusing on visual features or leveraging large-scale pre-training, most existing evaluations rely on rigid output formats and objective metrics, thus ignoring the complex, real-world demands of practical chart analysis.In this paper, we introduce ChartMind, a new benchmark designed for complex CQA tasks in real-world settings.ChartMind covers seven task categories, incorporates multilingual contexts, supports open-domain textual outputs, and accommodates diverse chart formats, bridging the gap between real-world applications and traditional academic benchmarks.Furthermore, we propose a context-aware yet modelagnostic framework, ChartLLM, that focuses on extracting key contextual elements, reducing noise, and enhancing the reasoning accuracy of multimodal large language models.Extensive evaluations on ChartMind and three representative public benchmarks with 14 mainstream multimodal models show our framework significantly outperforms the previous three common CQA paradigms: instruction-following, OCRenhanced, and chain-of-thought, highlighting the importance of flexible chart understanding for real-world CQA.These findings suggest new directions for developing more robust chart reasoning in future research.

Hand-drawn sketches are a natural and efficient medium for capturing and conveying ideas. Despite significant advancements in controllable natural image generation, translating freehand sketches into structured, machine-readable diagrams remains a labor-intensive and predominantly manual task. The primary challenge stems from the inherent ambiguity of sketches, which lack the structural constraints and semantic precision required for automated diagram generation. To address this challenge, we introduce SketchAgent, a multi-agent system designed to automate the transformation of hand-drawn sketches into structured diagrams. SketchAgent integrates sketch recognition, symbolic reasoning, and iterative validation to produce semantically coherent and structurally accurate diagrams, significantly reducing the need for manual effort. To evaluate the effectiveness of our approach, we propose the Sketch2Diagram Benchmark, a comprehensive dataset and evaluation framework encompassing eight diverse diagram categories, such as flowcharts, directed graphs, and model architectures. The dataset comprises over 6,000 high-quality examples with token-level annotations, standardized preprocessing, and rigorous quality control. By streamlining the diagram generation process, SketchAgent holds great promise for applications in design, education, and engineering, while offering a significant step toward bridging the gap between intuitive sketching and machine-readable diagram generation.

Understanding how scientific ideas evolve requires more than summarizing individual papers-it demands structured, cross-document reasoning over thematically related research. In this work, we formalize multi-document scientific inference, a new task that extracts and aligns motivation, methodology, and experimental results across related papers to reconstruct research development chains. This task introduces key challenges, including temporally aligning loosely structured methods and standardizing heterogeneous experimental tables. We present ResearchPulse, an agent-based framework that integrates instruction planning, scientific content extraction, and structured visualization. It consists of three coordinated agents: a Plan Agent for task decomposition, a Mmap-Agent that constructs motivation-method mind maps, and a Lchart-Agent that synthesizes experimental line charts. To support this task, we introduce ResearchPulse-Bench, a citation-aware benchmark of annotated paper clusters. Experiments show that our system, despite using 7B-scale agents, consistently outperforms strong baselines like GPT-4o in semantic alignment, structural consistency, and visual fidelity. The dataset are available in https://huggingface.co/datasets/ResearchPulse/ResearchPulse-Bench

Digital human generation is a forward-looking field in technology. Despite significant progress in the generation of speaking facial videos, many challenges remain unaddressed. Issues such as unnatural head movements, distorted expressions, artifacts in generated videos, and uncoordinated limb movements persist. Most current efforts are focused on specific individuals, with enhancements often limited to head movements without further advancing the overall behavioral actions of digital humans. In this context, we introduce a new dataset, CFMD, and a novel model, SAM-Wav2lip++, capable of generating consistent, audio-synchronized lip and behavior action videos from a single reference image of any identity. This work features three main innovative components: (1) a contrastive lip-sync discriminator for precise lip synchronization, (2) a generator for the synthesis of sound-action consistency, and (3) the SAM module for facial refinement operations. Through extensive experiments and user studies, our results demonstrate that our model can synthesize digital human videos of impressively high perceptual quality that accurately sync lip movements and behavioral actions with the input audio, substantially outperforming the state-of-the-art baselines evaluations.

In recent years, there has been significant progress in text-to-image generation models. However, text struggles to accurately describe abstract concepts like shapes and sizes. Some methods have been proposed to enhance text prompt by incorporating image prompts. While they have shown effective improvements, they either require substantial fine-tuning costs or struggle to effectively integrate text and image information. In our study, we delve into the issue of the difficulty in integrating text and image information in decoupled cross-attention and conduct visual analysis. We identify the presence of background-related tokens in image features as a key factor affecting text fidelity. To address this issue, we develop an algorithm to filter out these tokens. Additionally, we observe differences in the attention of Unet layers to text prompts and image prompts. Based on this finding, we optimize the flow of image information to reduce interference with text information. In summary, we introduce a new topic-customized method that requires no repeated training. It trains a plug-and-play image prompt adapter with only 417M parameters, lightweight yet powerful, surpassing existing models in both text and image consistency. Our code and pre-trained checkpoints will be available at https://github.com/YZBPXX/DDCA.

When solving constrained multi-objective optimization problems, the challenge is that how to deal with all kinds of constraints regardless of the shape of the feasible region. Especially when the feasible region is discrete or very small, some constraint handling techniques cannot solve it exactly. To address this issue, this paper proposes a new technique to handle constraints. First, all the constraints will be sorted to some grades from hard to easy according to their constrained violations. Second, a niching crowding distance mechanism is used to guarantee the diversity of the pareto front better. The experiments show that the proposed algorithm can generate a set uniformly distributed pareto optimal solutions under constrains.

Bi-level optimization problems with constraints are well known for wide applications and difficulties. A new particle swarm optimization algorithm (PSO) with dynamic constraint processing is proposed. First, we divide constraints into three cases and design a new PSO based crossover operator to dynamically deal with constraints; second, a normal distribution is added into PSO to work as mutation operator, which will enhance the diversity of the swarm and navigate the search direction. Finally, we give the convergence proof of the proposed algorithm.

The Suffix Array (SA) is a fundamental data structure which is widely used in the applications such as string matching, text index and computation biology, etc. How to sort the suffixes of a string in lexicographical order is a primary problem in constructing SAs, and one of the widely used suffix sorting algorithms is qsufsort. However, qsufsort suffers one critical limitation that the order of suffixes starting with the same [Formula: see text] characters cannot be determined in the kth round. To this point, in our paper, an efficient suffix sorting algorithm called dsufsort is proposed by overcoming the drawback of the qsufsort algorithm. In particular, our proposal maintains the depth of each unsorted portion of SA, and sorts the suffixes based on the depth in each round. By this means, some suffixes that cannot be sorted by qsufsort in each round can be sorted now, as a result, more sorting results in current round can be utilized by the latter rounds and the total number of sorting rounds will be reduced, which means dsufsort is more efficient than qsufsort. The experimental results show the effectiveness of the proposed algorithm, especially for the text with high repetitions.

Among the many-objective optimization problems, there exists a kind of problem with redundant objectives, it is possible to design effective algorithms by removing the redundant objectives and keeping the non-redundant objectives so that the original problem becomes the one with much fewer objectives. In this paper, a new non-redundant objective set generation algorithm is proposed. To do so, first, a multi-objective evolutionary algorithm based decomposition is adopted to generate a small number of representative non-dominated solutions widely distributed on the Pareto front. Then, the conflicting objective pairs are identified through these non-dominated solutions, and the non-redundant objective set is determined by these pairs. Finally, the experiments are conducted on a set of benchmark test problems and the results indicate the effectiveness and efficiency of the proposed algorithm.

The era of big data computing is coming. As scientific applications become more data intensive, finding an efficient scheduling strategy for massive computing in parallel and distributed systems has drawn increasingly attention. Most existing studies considered single-installment scheduling models, but very few literature involved multi-installment scheduling, especially in heterogeneous parallel and distributed systems. In this paper, we proposed a new model for periodic multi-installment divisible-load scheduling in which the make-span of the workload is minimized, and a genetic algorithm was designed to solve this model. Finally, experimental results show the effectiveness and efficiency of the proposed algorithm.

In the real world, many optimization problems are dynamic constrained multi-objective optimization problems. This requires an optimization algorithm not only to find the global optimal solutions under a specific environment but also to track the trajectory of the varying optima over dynamic environments. To address this requirement, this paper proposes a novel particle swarm optimization algorithm for such problems. This algorithm employs a new points selection strategy to speed up evolutionary process, and a local search operator to search optimal solutions in a promising subregion. The new algorithm is examined and compared with two well-known algorithms on a sequence of benchmark functions. The results show that the proposed algorithm can effectively track the varying Pareto fronts over time. The proposed developments are effective individually, but the combined effect is much better for the test functions.

In the real world, many optimization problems are dynamic constrained multi-objective optimization problems. This requires an optimization algorithm not only to find the global optimal solutions under a specific environment but also to track the trajectory of the varying optima over dynamic environments. To address this requirement, a hyper rectangle search based particle swarm algorithm is proposed for such problems. This algorithm employs a hyper rectangle search to predict the optimal solutions (in variable space) of the next time step. Then, a PSO based crossover operator is used to deal with all kinds of constraints appearing in the problems when the time step (environment) is fixed. This algorithm is tested and compared with two well known algorithms on a set of benchmarks. The results show that the proposed algorithm can effectively track the varying Pareto fronts over time.

In this paper, a new memetic algorithm for constrained multi-objective optimization problems is proposed, which combines the global search ability of particle swarm optimization with an attraction based local search operator for directed local fine-tuning. Firstly, a new particle updating strategy is proposed based on the concept of uncertain personal-best to deal with the problem of premature convergence. Secondly, an attraction based local search operator is proposed to find good local search direction for the particles. Finally, the convergence of the algorithm is proved. The proposed algorithm is examined and compared with two well known existing algorithms on five benchmark test functions. The results suggest that the new algorithm can evolve more good solutions, and the solutions are more widely spread and uniformly distributed along the Pareto front than the two existing methods. The proposed two developments are effective individually, but the combined effect is much better for these constrained multi-objective optimization problems.

In this paper, an infeasible elitist based particle swarm optimization is proposed for solving constrained optimization problems. Firstly, an infeasible elitist preservation strategy is proposed, which keeps some infeasible solutions with smaller rank values at the early stage of evolution regardless of how large the constraint violations are, and keep some infeasible solutions with smaller constraint violations and rank values at the later stage of evolution. In this manner, the true Pareto front will be found easier. Secondly, in order to find a set of diversity and uniformly distributed Pareto optimal solutions, a new crowding distance function is designed. It can assign large function values not only for the particles located in the sparse regions of the objective space but also for the crowded particles located near to the boundary of the Pareto front as well. Thirdly, a new mutation operator with two phases is proposed. In the first phase, the particles whose constraint violations are less than the threshold value will be used to compute the total force, then the force will be used as a mutation direction, being helpful to find the better solutions along this direction. In order to guarantee the convergence of the algorithm, the second phase of mutation is proposed. Finally, the convergence of the algorithm is proved. The comparative study shows that the proposed algorithm can generate widespread and uniformly distributed Pareto fronts and outperforms those compared algorithms.

Designing efficient algorithms for difficult multi-objective optimization problems is a very challenging problem. In this paper a new clustering multi-objective evolutionary algorithm based on orthogonal and uniform design is proposed. First, the orthogonal design is used to generate initial population of points that are scattered uniformly over the feasible solution space, so that the algorithm can evenly scan the feasible solution space once to locate good points for further exploration in subsequent iterations. Second, to explore the search space efficiently and get uniformly distributed and widely spread solutions in objective space, a new crossover operator is designed. Its exploration focus is mainly put on the sparse part and the boundary part of the obtained non-dominated solutions in objective space. Third, to get desired number of well distributed solutions in objective space, a new clustering method is proposed to select the non-dominated solutions. Finally, experiments on thirteen very difficult benchmark problems were made, and the results indicate the proposed algorithm is efficient.