Li Auto Inc.

Selected work

Representative Papers

MVGS: Multi-view-regulated Gaussian Splatting for Novel View Synthesis

Oct 02, 2024arXiv.org

To address overfitting in 3D Gaussian Splatting (3DGS) under single-view supervision—leading to artifacts in novel-view synthesis and inaccurate geometric reconstruction—this paper proposes a multi-view collaborative optimization framework. Our method introduces three key innovations: (1) a novel multi-view regularization paradigm that jointly enforces consistency across multiple views; (2) an intrinsic-cross-guided coarse-to-fine training strategy integrating multi-scale geometric and appearance priors; and (3) ray-intersection-driven cross-view densification coupled with view-difference-aware adaptive densification. While preserving real-time rendering performance, our approach significantly improves both novel-view image fidelity and 3D geometric accuracy. Extensive experiments demonstrate strong generalization across diverse scenes and mainstream 3DGS variants, outperforming existing single-view methods in both qualitative and quantitative evaluations.

15 citations1 influentialRead paper

Learning to Remember: End-to-End Training of Memory Agents for Long-Context Reasoning

Feb 13, 2026arXiv.org

为解决长上下文推理中状态跟踪不稳定的问题，本文提出统一记忆代理（UMA）通过一次构建查询无关的外部记忆并重复使用，结合任务分层GRPO方法进行训练。

6 citationsRead paper

RubricHub: A Comprehensive and Highly Discriminative Rubric Dataset via Automated Coarse-to-Fine Generation

Jan 13, 2026

This work addresses the challenge that open-ended generation tasks lack verifiable, fine-grained scoring rubrics, which limits the effectiveness of rule-based reinforcement learning. To overcome this, the authors propose an automated coarse-to-fine rubric generation framework that leverages principle-guided synthesis, multi-model aggregation, and a difficulty evolution mechanism to construct high-quality, highly discriminative scoring criteria. This framework enables the first large-scale, multi-domain, fine-grained, and scalable automatic evaluation system. Integrating the generated rubrics with Rejection Sampling Fine-Tuning (RuFT) and Rubric-guided Reinforcement Learning (RuRL), a Qwen3-14B model trained on RubricHub achieves a score of 69.3 on HealthBench, surpassing closed-source models such as GPT-5 and establishing a new state-of-the-art performance.

1 citationsRead paper

InfiniDepth: Arbitrary-Resolution and Fine-Grained Depth Estimation with Neural Implicit Fields

Jan 06, 2026arXiv.org

This work addresses the limitations of conventional depth estimation methods, which are constrained by discrete image grids and thus struggle to recover fine geometric details or support arbitrary-resolution outputs. To overcome this, we propose a continuous depth representation based on neural implicit fields, introducing a local implicit decoder that enables high-fidelity depth querying at any 2D coordinate. To facilitate training and evaluation, we construct a high-resolution 4K synthetic dataset. Experimental results demonstrate that our approach achieves state-of-the-art performance on both synthetic and real-world datasets, significantly enhancing geometric detail recovery and substantially improving the quality of novel-view synthesis under large viewpoint changes.

1 citationsRead paper