Kimi-VL Technical Report

We present Kimi-VL, an efficient open-source Mixture-of-Experts (MoE) vision-language model (VLM) that offers advanced multimodal reasoning, long-context understanding, and strong agent capabilities - all while activating only 2.8B parameters in its language decoder (Kimi-VL-A3B). Kimi-VL demonstrates strong performance across challenging domains: as a general-purpose VLM, Kimi-VL excels in multi-turn agent tasks (e.g., OSWorld), matching flagship models. Furthermore, it exhibits remarkable capabilities across diverse challenging vision language tasks, including college-level image and video comprehension, OCR, mathematical reasoning, and multi-image understanding. In comparative evaluations, it effectively competes with cutting-edge efficient VLMs such as GPT-4o-mini, Qwen2.5-VL-7B, and Gemma-3-12B-IT, while surpassing GPT-4o in several key domains. Kimi-VL also advances in processing long contexts and perceiving clearly. With a 128K extended context window, Kimi-VL can process diverse long inputs, achieving impressive scores of 64.5 on LongVideoBench and 35.1 on MMLongBench-Doc. Its native-resolution vision encoder, MoonViT, further allows it to see and understand ultra-high-resolution visual inputs, achieving 83.2 on InfoVQA and 34.5 on ScreenSpot-Pro, while maintaining lower computational cost for common tasks. Building upon Kimi-VL, we introduce an advanced long-thinking variant: Kimi-VL-Thinking. Developed through long chain-of-thought (CoT) supervised fine-tuning (SFT) and reinforcement learning (RL), this model exhibits strong long-horizon reasoning capabilities. It achieves scores of 61.7 on MMMU, 36.8 on MathVision, and 71.3 on MathVista while maintaining the compact 2.8B activated LLM parameters, setting a new standard for efficient multimodal thinking models. Code and models are publicly accessible at https://github.com/MoonshotAI/Kimi-VL.

SVG-IR: Spatially-Varying Gaussian Splatting for Inverse Rendering

Reconstructing 3D assets from images, known as inverse rendering (IR), remains a challenging task due to its ill-posed nature. 3D Gaussian Splatting (3DGS) has demonstrated impressive capabilities for novel view synthesis (NVS) tasks. Methods apply it to relighting by separating radiance into BRDF parameters and lighting, yet produce inferior relighting quality with artifacts and unnatural indirect illumination due to the limited capability of each Gaussian, which has constant material parameters and normal, alongside the absence of physical constraints for indirect lighting. In this paper, we present a novel framework called Spatially-vayring Gaussian Inverse Rendering (SVG-IR), aimed at enhancing both NVS and relighting quality. To this end, we propose a new representation-Spatially-varying Gaussian (SVG)-that allows per-Gaussian spatially varying parameters. This enhanced representation is complemented by a SVG splatting scheme akin to vertex/fragment shading in traditional graphics pipelines. Furthermore, we integrate a physically-based indirect lighting model, enabling more realistic relighting. The proposed SVG-IR framework significantly improves rendering quality, outperforming state-of-the-art NeRF-based methods by 2.5 dB in peak signal-to-noise ratio (PSNR) and surpassing existing Gaussian-based techniques by 3.5 dB in relighting tasks, all while maintaining a real-time rendering speed.

SSLFusion: Scale & Space Aligned Latent Fusion Model for Multimodal 3D Object Detection

Multimodal 3D object detection based on deep neural networks has indeed made significant progress. However, it still faces challenges due to the misalignment of scale and spatial information between features extracted from 2D images and those derived from 3D point clouds. Existing methods usually aggregate multimodal features at a single stage. However, leveraging multi-stage cross-modal features is crucial for detecting objects of various scales. Therefore, these methods often struggle to integrate features across different scales and modalities effectively, thereby restricting the accuracy of detection. Additionally, the time-consuming Query-Key-Value-based (QKV-based) cross-attention operations often utilized in existing methods aid in reasoning the location and existence of objects by capturing non-local contexts. However, this approach tends to increase computational complexity. To address these challenges, we present SSLFusion, a novel Scale & Space Aligned Latent Fusion Model, consisting of a scale-aligned fusion strategy (SAF), a 3D-to-2D space alignment module (SAM), and a latent cross-modal fusion module (LFM). SAF mitigates scale misalignment between modalities by aggregating features from both images and point clouds across multiple levels. SAM is designed to reduce the inter-modal gap between features from images and point clouds by incorporating 3D coordinate information into 2D image features. Additionally, LFM captures cross-modal non-local contexts in the latent space without utilizing the QKV-based attention operations, thus mitigating computational complexity. Experiments on the KITTI and DENSE datasets demonstrate that our SSLFusion outperforms state-of-the-art methods. Our approach obtains an absolute gain of 2.15% in 3D AP, compared with the state-of-art method GraphAlign on the moderate level of the KITTI test set.

Learning Class Prototypes for Unified Sparse Supervised 3D Object Detection

Both indoor and outdoor scene perceptions are essential for embodied intelligence. However, current sparse supervised 3D object detection methods focus solely on outdoor scenes without considering indoor settings. To this end, we propose a unified sparse supervised 3D object detection method for both indoor and outdoor scenes through learning class prototypes to effectively utilize unlabeled objects. Specifically, we first propose a prototype-based object mining module that converts the unlabeled object mining into a matching problem between class prototypes and unlabeled features. By using optimal transport matching results, we assign prototype labels to high-confidence features, thereby achieving the mining of unlabeled objects. We then present a multi-label cooperative refinement module to effectively recover missed detections through pseudo label quality control and prototype label cooperation. Experiments show that our method achieves state-of-the-art performance under the one object per scene sparse supervised setting across indoor and outdoor datasets. With only one labeled object per scene, our method achieves about 78%, 90%, and 96% performance compared to the fully supervised detector on ScanNet V2, SUN RGB-D, and KITTI, respectively, highlighting the scalability of our method. Code is available at https://github.com/zyrant/CPDet3D.

GigaSLAM: Large-Scale Monocular SLAM with Hierachical Gaussian Splats

Tracking and mapping in large-scale, unbounded outdoor environments using only monocular RGB input presents substantial challenges for existing SLAM systems. Traditional Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS) SLAM methods are typically limited to small, bounded indoor settings. To overcome these challenges, we introduce GigaSLAM, the first NeRF/3DGS-based SLAM framework for kilometer-scale outdoor environments, as demonstrated on the KITTI and KITTI 360 datasets. Our approach employs a hierarchical sparse voxel map representation, where Gaussians are decoded by neural networks at multiple levels of detail. This design enables efficient, scalable mapping and high-fidelity viewpoint rendering across expansive, unbounded scenes. For front-end tracking, GigaSLAM utilizes a metric depth model combined with epipolar geometry and PnP algorithms to accurately estimate poses, while incorporating a Bag-of-Words-based loop closure mechanism to maintain robust alignment over long trajectories. Consequently, GigaSLAM delivers high-precision tracking and visually faithful rendering on urban outdoor benchmarks, establishing a robust SLAM solution for large-scale, long-term scenarios, and significantly extending the applicability of Gaussian Splatting SLAM systems to unbounded outdoor environments.

Diff-Reg v2: Diffusion-Based Matching Matrix Estimation for Image Matching and 3D Registration

Establishing reliable correspondences is crucial for all registration tasks, including 2D image registration, 3D point cloud registration, and 2D-3D image-to-point cloud registration. However, these tasks are often complicated by challenges such as scale inconsistencies, symmetry, and large deformations, which can lead to ambiguous matches. Previous feature-based and correspondence-based methods typically rely on geometric or semantic features to generate or polish initial potential correspondences. Some methods typically leverage specific geometric priors, such as topological preservation, to devise diverse and innovative strategies tailored to a given enhancement goal, which cannot be exhaustively enumerated. Additionally, many previous approaches rely on a single-step prediction head, which can struggle with local minima in complex matching scenarios. To address these challenges, we introduce an innovative paradigm that leverages a diffusion model in matrix space for robust matching matrix estimation. Our model treats correspondence estimation as a denoising diffusion process in the matching matrix space, gradually refining the intermediate matching matrix to the optimal one. Specifically, we apply the diffusion model in the doubly stochastic matrix space for 3D-3D and 2D-3D registration tasks. In the 2D image registration task, we deploy the diffusion model in a matrix subspace where dual-softmax projection regularization is applied. For all three registration tasks, we provide adaptive matching matrix embedding implementations tailored to the specific characteristics of each task while maintaining a consistent "match-to-warp" encoding pattern. Furthermore, we adopt a lightweight design for the denoising module. In inference, once points or image features are extracted and fixed, this module performs multi-step denoising predictions through reverse sampling.

CLIPer: Hierarchically Improving Spatial Representation of CLIP for Open-Vocabulary Semantic Segmentation

Contrastive Language-Image Pre-training (CLIP) exhibits strong zero-shot classification ability on various image-level tasks, leading to the research to adapt CLIP for pixel-level open-vocabulary semantic segmentation without additional training. The key is to improve spatial representation of image-level CLIP, such as replacing self-attention map at last layer with self-self attention map or vision foundation model based attention map. In this paper, we present a novel hierarchical framework, named CLIPer, that hierarchically improves spatial representation of CLIP. The proposed CLIPer includes an early-layer fusion module and a fine-grained compensation module. We observe that, the embeddings and attention maps at early layers can preserve spatial structural information. Inspired by this, we design the early-layer fusion module to generate segmentation map with better spatial coherence. Afterwards, we employ a fine-grained compensation module to compensate the local details using the self-attention maps of diffusion model. We conduct the experiments on seven segmentation datasets. Our proposed CLIPer achieves the state-of-the-art performance on these datasets. For instance, using ViT-L, CLIPer has the mIoU of 69.8% and 43.3% on VOC and COCO Object, outperforming ProxyCLIP by 9.2% and 4.1% respectively.

Grid4D: 4D Decomposed Hash Encoding for High-fidelity Dynamic Gaussian Splatting

Recently, Gaussian splatting has received more and more attention in the field of static scene rendering. Due to the low computational overhead and inherent flexibility of explicit representations, plane-based explicit methods are popular ways to predict deformations for Gaussian-based dynamic scene rendering models. However, plane-based methods rely on the inappropriate low-rank assumption and excessively decompose the space-time 4D encoding, resulting in overmuch feature overlap and unsatisfactory rendering quality. To tackle these problems, we propose Grid4D, a dynamic scene rendering model based on Gaussian splatting and employing a novel explicit encoding method for the 4D input through the hash encoding. Different from plane-based explicit representations, we decompose the 4D encoding into one spatial and three temporal 3D hash encodings without the low-rank assumption. Additionally, we design a novel attention module that generates the attention scores in a directional range to aggregate the spatial and temporal features. The directional attention enables Grid4D to more accurately fit the diverse deformations across distinct scene components based on the spatial encoded features. Moreover, to mitigate the inherent lack of smoothness in explicit representation methods, we introduce a smooth regularization term that keeps our model from the chaos of deformation prediction. Our experiments demonstrate that Grid4D significantly outperforms the state-of-the-art models in visual quality and rendering speed.

FastPCI: Motion-Structure Guided Fast Point Cloud Frame Interpolation

Point cloud frame interpolation is a challenging task that involves accurate scene flow estimation across frames and maintaining the geometry structure. Prevailing techniques often rely on pre-trained motion estimators or intensive testing-time optimization, resulting in compromised interpolation accuracy or prolonged inference. This work presents FastPCI that introduces Pyramid Convolution-Transformer architecture for point cloud frame interpolation. Our hybrid Convolution-Transformer improves the local and long-range feature learning, while the pyramid network offers multilevel features and reduces the computation. In addition, FastPCI proposes a unique Dual-Direction Motion-Structure block for more accurate scene flow estimation. Our design is motivated by two facts: (1) accurate scene flow preserves 3D structure, and (2) point cloud at the previous timestep should be reconstructable using reverse motion from future timestep. Extensive experiments show that FastPCI significantly outperforms the state-of-the-art PointINet and NeuralPCI with notable gains (e.g. 26.6% and 18.3% reduction in Chamfer Distance in KITTI), while being more than 10x and 600x faster, respectively. Code is available at https://github.com/genuszty/FastPCI

iSeg: An Iterative Refinement-based Framework for Training-free Segmentation

Stable diffusion has demonstrated strong image synthesis ability to given text descriptions, suggesting it to contain strong semantic clue for grouping objects. Inspired by this, researchers have explored employing stable diffusion for trainingfree segmentation. Most existing approaches either simply employ cross-attention map or refine it by self-attention map, to generate segmentation masks. We believe that iterative refinement with self-attention map would lead to better results. However, we mpirically demonstrate that such a refinement is sub-optimal likely due to the self-attention map containing irrelevant global information which hampers accurately refining cross-attention map with multiple iterations. To address this, we propose an iterative refinement framework for training-free segmentation, named iSeg, having an entropy-reduced self-attention module which utilizes a gradient descent scheme to reduce the entropy of self-attention map, thereby suppressing the weak responses corresponding to irrelevant global information. Leveraging the entropy-reduced self-attention module, our iSeg stably improves refined crossattention map with iterative refinement. Further, we design a category-enhanced cross-attention module to generate accurate cross-attention map, providing a better initial input for iterative refinement. Extensive experiments across different datasets and diverse segmentation tasks reveal the merits of proposed contributions, leading to promising performance on diverse segmentation tasks. For unsupervised semantic segmentation on Cityscapes, our iSeg achieves an absolute gain of 3.8% in terms of mIoU compared to the best existing training-free approach in literature. Moreover, our proposed iSeg can support segmentation with different kind of images and interactions.