Qihang Ai, Pi Bu, Yue Cao, Yingyao Wang, Jihao Gu, Jingxuan Xing, Zekun Zhu, Wei Jiang, Zhicheng Zheng, Jun Song, Yuning Jiang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Yancheng He, Shilong Li, Jiaheng Liu, Yingshui Tan, Weixun Wang, Hui Huang, Xingyuan Bu, Hangyu Guo, Chengwei Hu, Boren Zheng, Zhuoran Lin, Dekai Sun, Zhicheng Zheng, Wenbo Su, Bo Zheng. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Yancheng He, Shilong Li, Jiaheng Liu, Weixun Wang, Xingyuan Bu, Ge Zhang, Z.y. Peng, Zhaoxiang Zhang, Zhicheng Zheng, Wenbo Su, Bo Zheng. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Recent advances in text-to-video (T2V) generation highlight the critical role of highquality video-text pairs in training models capable of producing coherent and instructionaligned videos.However, strategies for optimizing video captions specifically for T2V training remain underexplored.In this paper, we introduce VC4VG (Video Captioning for Video Generation), a comprehensive caption optimization framework tailored to the needs of T2V models.We begin by analyzing caption content from a T2V perspective, decomposing the essential elements required for video reconstruction into multiple dimensions, and proposing a principled caption design methodology.To support evaluation, we construct VC4VG-Bench, a new benchmark featuring fine-grained, multi-dimensional, and necessity-graded metrics aligned with T2Vspecific requirements.Extensive T2V finetuning experiments demonstrate a strong correlation between improved caption quality and video generation performance, validating the effectiveness of our approach.We release all benchmark tools and code 1 to support further research.

We develop a new technique for proving distribution testing lower bounds for properties defined by inequalities on the individual bin probabilities (such as monotonicity and log-concavity). The basic idea is to find a base distribution $Q$ where these inequalities barely hold in many places. We then find two different ensembles of distributions that modify $Q$ in slightly different ways. We use a moment matching construction so that each ensemble has the same bin moments (in particular the expectation over the choice of distribution $p$ of $p_{i}^t$ is the same for the two ensembles for small integers $t$). We show that this makes it impossible to distinguish between the two ensembles with a small number of samples. On the other hand, we construct them so that one ensemble will tweak Q in such a way that it may violate the defining inequalities of the property in question in many places, while the second ensembles does not. Since any valid tester for this property must be able to reliably distinguish these ensembles, we obtain a lower bound of testing the property. Roughly speaking, if we can construct Q which nearly violates the defining inequalities in n places and if the desired error $\epilon$ is small enough relative to n, we hope to obtain a lower bound of roughly $\frac{n}{\epsilon^2}$ up to log factors. In particular, we obtain a lower bound of $\Omega( \min(n,(1/\epsilon)/ \log^3(1/\epsilon))\allowbreak / ( \epsilon^2 \log^7(1/\epsilon)))$ for monotonicity testing on $[n]$ and $\Omega(\log^{-7}(1/\epsilon) \epsilon^{-2} \min(n,\epsilon^{-1/2}\log^{-3/2}(1/\epsilon)))$ for log-concavity testing on $[n]$, the latter of which matches known upper bounds to within logarithmic factors. More generally, for monotonicity testing on $[n]^d$, we have the lower bound of $2^{-O(d)}d^{-d} \epsilon^{-2} \log^{-7}(1/\epsilon) \min(n,d \epsilon^{-1} \log^{-3}(1/\epsilon))^d$.

We introduce the Continuum Physical Dataset (ContPhy), a novel benchmark for assessing machine physical commonsense. ContPhy complements existing physical reasoning benchmarks by encompassing the inference of diverse physical properties, such as mass and density, across various scenarios and predicting corresponding dynamics. We evaluated a range of AI models and found that they still struggle to achieve satisfactory performance on ContPhy, which shows that current AI models still lack physical commonsense for the continuum, especially soft-bodies, and illustrates the value of the proposed dataset. We also introduce an oracle model (ContPRO) that marries the particle-based physical dynamic models with the recent large language models, which enjoy the advantages of both models, precise dynamic predictions, and interpretable reasoning. ContPhy aims to spur progress in perception and reasoning within diverse physical settings, narrowing the divide between human and machine intelligence in understanding the physical world.

A swarm of robots is a typical self-organizing system. Parameters of self-organizing model can regulate the interaction among the individuals to achieve the desired collective motion. Thus, it makes sense to optimize the parameters for better performance in self-organization. The evolutionary algorithm (EA) is one promising heuristic optimization algorithm. However, quantitative evaluation of a self-organizing model is still an open scientific question. In this paper, we proposed three metrics for collective motion from the perspective of grouping, polarization and spatial distribution of group. Then, we designed the fitness function with the proposed metrics to bridge the gap between EA and parameter tuning of self-organizing model. In the experimental study, we optimized two kinds of self-organizing models using the differential evolution algorithm (DE) and validated the effectiveness in simulation. Moreover, we validated the optimized parameters on a swarm of physical robots. The experimental results show that the proposed metrics are effective to quantitatively evaluate the self-organizing model so that the DE performs well for the automatic parameter tuning of controlling swarm robots.

Entity disambiguation with a knowledge base becomes increasingly popular in the NLP community. In this paper, we employ Freebase as the knowledge base, which contains significantly more entities than Wikipedia and others. While huge in size, Freebase lacks context for most entities, such as the descriptive text and hyperlinks in Wikipedia, which are useful for disambiguation. Instead, we leverage two features of Freebase, namely the naturally disambiguated mention phrases (aka aliases) and the rich taxonomy, to perform disambiguation in an iterative manner. Specifically, we explore both generative and discriminative models for each iteration. Experiments on 2, 430, 707 English sentences and 33, 743 Freebase entities show the effectiveness of the two features, where 90% accuracy can be reached without any labeled data. We also show that discriminative models with proposed split training strategy is robust against over fitting problem, and constantly outperforms the generative ones.