Yuqin Yang, Haowu Zhou, Haoran Tu, Zhiwen Hui, Shiqi Yan, HaoYang Li, Dong She, Xianrong Yao, Yang Gao, Zhanpeng Jin. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Light detection and ranging (LiDAR) point cloud denoising is critical for reliable environmental perception in autonomous driving and robotics. To overcome the lack of real-noise datasets and the limited generalization of algorithms that rely on synthetic data, we construct a real-world LiDAR denoising dataset with noise-clean pairs, named RealLiD. Meanwhile, we propose a dual-heterogeneous-encoder network (DHE-Net) tailored for real-world noise. DHE-Net leverages spatial order information obtained from Knearest neighbor (KNN) sampling. It employs heterogeneous dual encoders to extract both the central semantic details and boundary distribution features of point cloud patches, thereby enabling more effective denoising. Experiments on RealLiD demonstrate that DHE-Net substantially outperforms mainstream denoising algorithms across multiple metrics, including chamfer distance, thereby proving its robustness and practicality under real-world noise conditions. The dataset and code will be released as open-source after publication to support future research.

In recent years, low-cost LiDAR has gained attention for its cost-effectiveness, but the noisy point cloud data it captures limits algorithm performance. We propose the R2R-LPCD dataset, a real-world LiDAR-based point cloud denoising dataset designed to address the limitations of synthetic data in capturing complex noise patterns. Comprising 162 high-quality sample pairs, R2R-LPCD uniquely reflects real-world noise characteristics, such as ray-like noise at object boundaries and occlusion-induced structural gaps, offering a robust platform for algorithm evaluation under practical conditions. This dataset supports advancements in sensor systems, embedded AI, and real-world applications by providing tools and benchmarks for resource-efficient machine learning and edge computing. By publicly releasing R2R-LPCD, we aim to drive innovation in low-cost LiDAR applications, particularly in autonomous driving and robotics, while addressing current technical challenges through future scalability and methodological improvements.

We study the problem of identifying the unknown intervention targets in structural causal models where we have access to heterogeneous data collected from multiple environments. The unknown intervention targets are the set of endogenous variables whose corresponding exogenous noises change across the environments. We propose a two-phase approach which in the first phase recovers the exogenous noises corresponding to unknown intervention targets whose distributions have changed across environments. In the second phase, the recovered noises are matched with the corresponding endogenous variables. For the recovery phase, we provide sufficient conditions for learning these exogenous noises up to some component-wise invertible transformation. For the matching phase, under the causal sufficiency assumption, we show that the proposed method uniquely identifies the intervention targets. In the presence of latent confounders, the intervention targets among the observed variables cannot be determined uniquely. We provide a candidate intervention target set which is a superset of the true intervention targets. Our approach improves upon the state of the art as the returned candidate set is always a subset of the target set returned by previous work. Moreover, we do not require restrictive assumptions such as linearity of the causal model or performing invariance tests to learn whether a distribution is changing across environments which could be highly sample inefficient. Our experimental results show the effectiveness of our proposed algorithm in practice.

We focus on causal discovery in the presence of measurement error in linear systems where the mixing matrix, i.e., the matrix indicating the independent exogenous noise terms pertaining to the observed variables, is identified up to permutation and scaling of the columns. We demonstrate a somewhat surprising connection between this problem and causal discovery in the presence of unobserved parentless causes, in the sense that there is a mapping, given by the mixing matrix, between the underlying models to be inferred in these problems. Consequently, any identifiability result based on the mixing matrix for one model translates to an identifiability result for the other model. We characterize to what extent the causal models can be identified under a two-part faithfulness assumption. Under only the first part of the assumption (corresponding to the conventional definition of faithfulness), the structure can be learned up to the causal ordering among an ordered grouping of the variables but not all the edges across the groups can be identified. We further show that if both parts of the faithfulness assumption are imposed, the structure can be learned up to a more refined ordered grouping. As a result of this refinement, for the latent variable model with unobserved parentless causes, the structure can be identified. Based on our theoretical results, we propose causal structure learning methods for both models, and evaluate their performance on synthetic data.