Fairness concerns are increasingly critical as machine learning models are deployed in high-stakes applications. While existing fairness-aware methods typically intervene at the model level, they often suffer from high computational costs, limited scalability, and poor generalization. To address these challenges, we propose a Bayesian data selection framework that ensures fairness by aligning group-specific posterior distributions of model parameters and sample weights with a shared central distribution. Our framework supports flexible alignment via various distributional discrepancy measures, including Wasserstein distance, maximum mean discrepancy, and f-divergence, allowing geometry-aware control without imposing explicit fairness constraints. This data-centric approach mitigates group-specific biases in training data and improves fairness in downstream tasks, with theoretical guarantees. Experiments on benchmark datasets show that our method consistently outperforms existing data selection and model-based fairness methods in both fairness and accuracy.

Quyu Kong, Xu Zhang, Zhenyu Yang, Nolan Gao, Chen Liu, Panrong Tong, Chenglin Cai, Hanzhang Zhou, Jianan Zhang, Liangyu Chen, Zhidan Liu, Steven Hoi, Yue Wang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Temporal Point Processes (TPPs) have been widely used for modeling event sequences on the Web, such as user reviews, social media posts, and online transactions. However, traditional TPP models often struggle to effectively incorporate the rich textual descriptions that accompany these events, while Large Language Models (LLMs), despite their remarkable text processing capabilities, lack mechanisms for handling the temporal dynamics inherent in Web-based event sequences. To bridge this gap, we introduce Language-TPP, a unified framework that seamlessly integrates TPPs with LLMs for enhanced Web event sequence modeling. Our key innovation is a novel temporal encoding mechanism that converts continuous time intervals into specialized byte-tokens, enabling direct integration with standard language model architectures for TPP modeling without requiring TPP-specific modifications. This approach allows Language-TPP to achieve state-of-the-art performance across multiple TPP benchmarks, including event time prediction and type prediction, on real-world Web datasets spanning e-commerce reviews, social media and online Q&A platforms. More importantly, we demonstrate that our unified framework unlocks new capabilities for TPP research: incorporating temporal information improves the quality of generated event descriptions, as evidenced by enhanced ROUGE-L scores, and better aligned sentiment distributions. Through comprehensive experiments, including qualitative analysis of learned distributions and scalability evaluations on long sequences, we show that Language-TPP effectively captures both temporal dynamics and textual patterns in Web user behavior, with important implications for content generation, user behavior understanding, and Web platform applications. Code is available at https://github.com/qykong/Language-TPP.

Grounding 3D object affordance is a task that locates objects in 3D space where they can be manipulated, which links perception and action for embodied intelligence. For example, for an intelligent robot, it is necessary to accurately ground the affordance of an object and grasp it according to human instructions. In this paper, we introduce a novel task that grounds 3D object affordance based on language instructions, visual observations and interactions, which is inspired by cognitive science. We collect an Affordance Grounding dataset with Points, Images and Language instructions (AGPIL) to support the proposed task. In the 3D physical world, due to observation orientation, object rotation, or spatial occlusion, we can only get a partial observation of the object. So this dataset includes affordance estimations of objects from full-view, partial-view, and rotation-view perspectives. To accomplish this task, we propose LMAffordance3D, the first multi-modal, language-guided 3D affordance grounding network, which applies a vision-language model to fuse 2D and 3D spatial features with semantic features. Comprehensive experiments on AGPIL demonstrate the effectiveness and superiority of our method on this task, even in unseen experimental settings. Our project is available at https://sites.google.com/view/lmaffordance3d.

We propose TPP-SD, a novel approach that accelerates Transformer temporal point process (TPP) sampling by adapting speculative decoding (SD) techniques from language models. By identifying the structural similarities between thinning algorithms for TPPs and speculative decoding for language models, we develop an efficient sampling framework that leverages a smaller draft model to generate multiple candidate events, which are then verified by the larger target model. TPP-SD maintains the same output distribution as autoregressive sampling while achieving significant acceleration. Experiments on both synthetic and real datasets demonstrate that our approach produces samples from identical distributions as standard methods, but with 2-6$\times$ speedup. Our ablation studies analyze the impact of hyperparameters such as draft length and draft model size on sampling efficiency. TPP-SD bridges the gap between powerful Transformer TPP models and the practical need for rapid sequence generation.

We introduce DanmakuTPPBench, a comprehensive benchmark designed to advance multi-modal Temporal Point Process (TPP) modeling in the era of Large Language Models (LLMs). While TPPs have been widely studied for modeling temporal event sequences, existing datasets are predominantly unimodal, hindering progress in models that require joint reasoning over temporal, textual, and visual information. To address this gap, DanmakuTPPBench comprises two complementary components:(1) DanmakuTPP-Events, a novel dataset derived from the Bilibili video platform, where user-generated bullet comments (Danmaku) naturally form multi-modal events annotated with precise timestamps, rich textual content, and corresponding video frames;(2) DanmakuTPP-QA, a challenging question-answering dataset constructed via a novel multi-agent pipeline powered by state-of-the-art LLMs and multi-modal LLMs (MLLMs), targeting complex temporal-textual-visual reasoning. We conduct extensive evaluations using both classical TPP models and recent MLLMs, revealing significant performance gaps and limitations in current methods’ ability to model multi-modal event dynamics. Our benchmark establishes strong baselines and calls for further integration of TPP modeling into the multi-modal language modeling landscape. Project page: https://github.com/FRENKIE-CHIANG/DanmakuTPPBench.

Vehicle-to-infrastructure cooperative 3D object detection (VIC3D) is a task that leverages both vehicle and roadside sensors to jointly perceive the surrounding environment. However, considering the high speed of vehicles, the real-time requirements, and the limitations of communication bandwidth, roadside devices transmit the results of perception rather than raw sensor data or feature maps in our real-world scenarios. And affected by various environmental factors, the transmission delay is dynamic. To meet the needs of practical applications, we present OTVIC, which is the first multi-modality and multi-view dataset with online transmission from real scenes for vehicle-to-infrastructure cooperative 3D object detection. The ego-vehicle receives the results of infrastructure perception in real-time, collected from a section of highway in Chengdu, China. Moreover, we propose LfFormer, which is a novel end-to-end multi-modality late fusion framework with transformer for VIC3D task as a baseline based on OTVIC. Experiments prove our fusion framework’s effectiveness and robustness. Our project is available at https://sites.google.com/view/otvic.

In this study, we propose a novel deep spatio-temporal point process model, Deep Kernel Mixture Point Processes (DKMPP), that incorporates multimodal covariate information. DKMPP is an enhanced version of Deep Mixture Point Processes (DMPP), which uses a more flexible deep kernel to model complex relationships between events and covariate data, improving the model's expressiveness. To address the intractable training procedure of DKMPP due to the non-integrable deep kernel, we utilize an integration-free method based on score matching, and further improve efficiency by adopting a scalable denoising score matching method. Our experiments demonstrate that DKMPP and its corresponding score-based estimators outperform baseline models, showcasing the advantages of incorporating covariate information, utilizing a deep kernel, and employing score-based estimators.

Social media is being increasingly weaponized by state-backed actors to elicit reactions, push narratives and sway public opinion. These are known as Information Operations (IO). The covert nature of IO makes their detection difficult. This is further amplified by missing data due to the user and content removal and privacy requirements. This work advances the hypothesis that the very reactions that Information Operations seek to elicit within the target social systems can be used to detect them. We propose an Interval-censored Transformer Hawkes (IC-TH) architecture and a novel data encoding scheme to account for both observed and missing data. We derive a novel log-likelihood function that we deploy together with a contrastive learning procedure. We showcase the performance of IC-TH on three real-world Twitter datasets and two learning tasks: future popularity prediction and item category prediction. The latter is particularly significant. Using the retweeting timing and patterns solely, we can predict the category of YouTube videos, guess whether news publishers are reputable or controversial and, most importantly, identify state-backed IO agent accounts. Additional qualitative investigations uncover that the automatically discovered clusters of Russian-backed agents appear to coordinate their behavior, activating simultaneously to push specific narratives.

Hawkes process provides an effective statistical framework for analyzing the interactions of neural spiking activities. Although utilized in many real applications, the classic Hawkes process is incapable of modeling inhibitory interactions among neural population. Instead, the nonlinear Hawkes process allows for modeling a more flexible influence pattern with excitatory or inhibitory interactions. This work proposes a flexible nonlinear Hawkes process variant based on sigmoid nonlinearity. To ease inference, three sets of auxiliary latent variables (Polya-Gamma variables, latent marked Poisson processes and sparsity variables) are augmented to make functional connection weights appear in a Gaussian form, which enables simple iterative algorithms with analytical updates. As a result, the efficient Gibbs sampler, expectation-maximization algorithm and mean-field approximation are derived to estimate the interactions among neural populations. Furthermore, to reconcile with time-varying neural systems, the proposed time-invariant model is extended to a dynamic version by introducing a Markov state process. Similarly, three analytical iterative inference algorithms: Gibbs sampler, EM algorithm and mean-field approximation are derived. We compare the accuracy and efficiency of these inference algorithms on synthetic data, and further experiment on real neural recordings to demonstrate that the developed models achieve superior performance over the state-of-the-art competitors.

Modeling online discourse dynamics is a core activity in understanding the spread of information, both offline and online, and emergent online behavior. There is currently a disconnect between the practitioners of online social media analysis --- usually social, political and communication scientists --- and the accessibility to tools capable of examining online discussions of users. Here we present evently, a tool for modeling online reshare cascades, and particularly retweet cascades, using self-exciting processes. It provides a comprehensive set of functionalities for processing raw data from Twitter public APIs, modeling the temporal dynamics of processed retweet cascades and characterizing online users with a wide range of diffusion measures. This tool is designed for researchers with a wide range of computer expertise, and it includes tutorials and detailed documentation. We illustrate the usage of evently with an end-to-end analysis of online user behavior on a topical dataset relating to COVID-19. We show that, by characterizing users solely based on how their content spreads online, we can disentangle influential users and online bots.

The impact of online social media on societal events and institutions is profound, and with the rapid increases in user uptake, we are just starting to understand its ramifications. Social scientists and practitioners who model online discourse as a proxy for real-world behavior often curate large social media datasets. A lack of available tooling aimed at non-data science experts frequently leaves this data (and the insights it holds) underutilized. Here, we propose birdspotter -- a tool to analyze and label Twitter users --, and birdspotter.ml -- an exploratory visualizer for the computed metrics. birdspotter provides an end-to-end analysis pipeline, from the processing of pre-collected Twitter data to general-purpose labeling of users and estimating their social influence, within a few lines of code. The package features tutorials and detailed documentation. We also illustrate how to train birdspotter into a fully-fledged bot detector that achieves better than state-of-the-art performances without making Twitter API calls, and we showcase its usage in an exploratory analysis of a topical COVID-19 dataset.

It is well-known that online behavior is long-tailed, with most cascaded actions being short and a few being very long. A prominent drawback in generative models for online events is the inability to describe unpopular items well. This work addresses these shortcomings by proposing dual mixture self-exciting processes to jointly learn from groups of cascades. We first start from the observation that maximum likelihood estimates for content virality and influence decay are separable in a Hawkes process. Next, our proposed model, which leverages a Borel mixture model and a kernel mixture model, jointly models the unfolding of a heterogeneous set of cascades. When applied to cascades of the same online items, the model directly characterizes their spread dynamics and supplies interpretable quantities, such as content virality and content influence decay, as well as methods for predicting the final content popularities. On two retweet cascade datasets --- one relating to YouTube videos and the second relating to controversial news articles --- we show that our models capture the differences between online items at the granularity of items, publishers and categories. In particular, we are able to distinguish between far-right, conspiracy, controversial and reputable online news articles based on how they diffuse through social media, achieving an F1 score of 0.945. On holdout datasets, we show that the dual mixture model provides, for reshare diffusion cascades especially unpopular ones, better generalization performance and, for online items, accurate item popularity predictions.

Epidemic models and self-exciting processes are two types of models used to describe diffusion phenomena online and offline. These models were originally developed in different scientific communities, and their commonalities are under-explored. This work establishes, for the first time, a general connection between the two model classes via three new mathematical components. The first is a generalized version of stochastic Susceptible-Infected-Recovered (SIR) model with arbitrary recovery time distributions; the second is the relationship between the (latent and arbitrary) recovery time distribution, recovery hazard function, and the infection kernel of self-exciting processes; the third includes methods for simulating, fitting, evaluating and predicting the generalized process. On three large Twitter diffusion datasets, we conduct goodness-of-fit tests and holdout log-likelihood evaluation of self-exciting processes with three infection kernels --- exponential, power-law and Tsallis Q-exponential. We show that the modeling performance of the infection kernels varies with respect to the temporal structures of diffusions, and also with respect to user behavior, such as the likelihood of being bots. We further improve the prediction of popularity by combining two models that are identified as complementary by the goodness-of-fit tests.

Epidemic models and Hawkes point process models are two common model classes for information diffusion. Recent work has revealed the equivalence between the two for information diffusion modeling. This allows tools created for one class of models to be applied to another. However, epidemic models and Hawkes point processes can be connected in more ways. This thesis aims to develop a rich set of mathematical equivalences and extensions, and use them to ask and answer questions in social media and beyond. Specifically, we show our plan of generalizing the equivalence of the two model classes by extending it to Hawkes point process models with arbitrary memory kernels. We then outline a rich set of quantities describing diffusion, including diffusion size and extinction probability, introduced in the fields where the models are originally designed. Lastly, we discuss some novel applications of these quantities in a range of problems such as popularity prediction and popularity intervention.

Among the statistical tools for online information diffusion modeling, both epidemic models and Hawkes point processes are popular choices. The former originate from epidemiology, and consider information as a viral contagion which spreads into a population of online users. The latter have roots in geophysics and finance, view individual actions as discrete events in continuous time, and modulate the rate of events according to the self-exciting nature of event sequences. Here, we establish a novel connection between these two frameworks. Namely, the rate of events in an extended Hawkes model is identical to the rate of new infections in the Susceptible-Infected-Recovered (SIR) model after marginalizing out recovery events -- which are unobserved in a Hawkes process. This result paves the way to apply tools developed for SIR to Hawkes, and vice versa. It also leads to HawkesN, a generalization of the Hawkes model which accounts for a finite population size. Finally, we derive the distribution of cascade sizes for HawkesN, inspired by methods in stochastic SIR. Such distributions provide nuanced explanations to the general unpredictability of popularity: the distribution for diffusion cascade sizes tends to have two modes, one corresponding to large cascade sizes and another one around zero.