JoyAI-RA 0.1: A Foundation Model for Robotic Autonomy

Robotic autonomy in open-world environments is fundamentally limited by insufficient data diversity and poor cross-embodiment generalization. Existing robotic datasets are often limited in scale and task coverage, while relatively large differences across robot embodiments impede effective behavior knowledge transfer. To address these challenges, we propose JoyAI-RA, a vision-language-action (VLA) embodied foundation model tailored for generalizable robotic manipulation. JoyAI-RA presents a multi-source multi-level pretraining framework that integrates web data, large-scale egocentric human manipulation videos, simulation-generated trajectories, and real-robot data. Through training on heterogeneous multi-source data with explicit action-space unification, JoyAI-RA effectively bridges embodiment gaps, particularly between human manipulation and robotic control, thereby enhancing cross-embodiment behavior learning. JoyAI-RA outperforms state-of-the-art methods in both simulation and real-world benchmarks, especially on diverse tasks with generalization demands.

JoyAI-LLM Flash: Advancing Mid-Scale LLMs with Token Efficiency

We introduce JoyAI-LLM Flash, an efficient Mixture-of-Experts (MoE) language model designed to redefine the trade-off between strong performance and token efficiency in the sub-50B parameter regime. JoyAI-LLM Flash is pretrained on a massive corpus of 20 trillion tokens and further optimized through a rigorous post-training pipeline, including supervised fine-tuning (SFT), Direct Preference Optimization (DPO), and large-scale reinforcement learning (RL) across diverse environments. To improve token efficiency, JoyAI-LLM Flash strategically balances \emph{thinking} and \emph{non-thinking} cognitive modes and introduces FiberPO, a novel RL algorithm inspired by fibration theory that decomposes trust-region maintenance into global and local components, providing unified multi-scale stability control for LLM policy optimization. To enhance architectural sparsity, the model comprises 48B total parameters while activating only 2.7B parameters per forward pass, achieving a substantially higher sparsity ratio than contemporary industry leading models of comparable scale. To further improve inference throughput, we adopt a joint training-inference co-design that incorporates dense Multi-Token Prediction (MTP) and Quantization-Aware Training (QAT). We release the checkpoints for both JoyAI-LLM-48B-A3B Base and its post-trained variants on Hugging Face to support the open-source community.

A Multi-modal Detection System for Infrastructure-based Freight Signal Priority

Freight vehicles approaching signalized intersections require reliable detection and motion estimation to support infrastructure-based Freight Signal Priority (FSP). Accurate and timely perception of vehicle type, position, and speed is essential for enabling effective priority control strategies. This paper presents the design, deployment, and evaluation of an infrastructure-based multi-modal freight vehicle detection system integrating LiDAR and camera sensors. A hybrid sensing architecture is adopted, consisting of an intersection-mounted subsystem and a midblock subsystem, connected via wireless communication for synchronized data transmission. The perception pipeline incorporates both clustering-based and deep learning-based detection methods with Kalman filter tracking to achieve stable real-time performance. LiDAR measurements are registered into geodetic reference frames to support lane-level localization and consistent vehicle tracking. Field evaluations demonstrate that the system can reliably monitor freight vehicle movements at high spatio-temporal resolution. The design and deployment provide practical insights for developing infrastructure-based sensing systems to support FSP applications.

What Really Matters for Robust Multi-Sensor HD Map Construction?

High-definition (HD) map construction methods are crucial for providing precise and comprehensive static environmental information, which is essential for autonomous driving systems. While Camera-LiDAR fusion techniques have shown promising results by integrating data from both modalities, existing approaches primarily focus on improving model accuracy and often neglect the robustness of perception models, which is a critical aspect for real-world applications. In this paper, we explore strategies to enhance the robustness of multi-modal fusion methods for HD map construction while maintaining high accuracy. We propose three key components: data augmentation, a novel multi-modal fusion module, and a modality dropout training strategy. These components are evaluated on a challenging dataset containing 10 days of NuScenes data. Our experimental results demonstrate that our proposed methods significantly enhance the robustness of baseline methods. Furthermore, our approach achieves state-of-the-art performance on the clean validation set of the NuScenes dataset. Our findings provide valuable insights for developing more robust and reliable HD map construction models, advancing their applicability in real-world autonomous driving scenarios. Project website: https://robomap-123.github.io.

VTLA: Vision-Tactile-Language-Action Model with Preference Learning for Insertion Manipulation

While vision-language models have advanced significantly, their application in language-conditioned robotic manipulation is still underexplored, especially for contact-rich tasks that extend beyond visually dominant pick-and-place scenarios. To bridge this gap, we introduce Vision-Tactile-Language-Action model, a novel framework that enables robust policy generation in contact-intensive scenarios by effectively integrating visual and tactile inputs through cross-modal language grounding. A low-cost, multi-modal dataset has been constructed in a simulation environment, containing vision-tactile-action-instruction pairs specifically designed for the fingertip insertion task. Furthermore, we introduce Direct Preference Optimization (DPO) to offer regression-like supervision for the VTLA model, effectively bridging the gap between classification-based next token prediction loss and continuous robotic tasks. Experimental results show that the VTLA model outperforms traditional imitation learning methods (e.g., diffusion policies) and existing multi-modal baselines (TLA/VLA), achieving over 90% success rates on unseen peg shapes. Finally, we conduct real-world peg-in-hole experiments to demonstrate the exceptional Sim2Real performance of the proposed VTLA model. For supplementary videos and results, please visit our project website: https://sites.google.com/view/vtla

CLTP: Contrastive Language-Tactile Pre-training for 3D Contact Geometry Understanding

Recent advancements in integrating tactile sensing with vision-language models (VLMs) have demonstrated remarkable potential for robotic multimodal perception. However, existing tactile descriptions remain limited to superficial attributes like texture, neglecting critical contact states essential for robotic manipulation. To bridge this gap, we propose CLTP, an intuitive and effective language tactile pretraining framework that aligns tactile 3D point clouds with natural language in various contact scenarios, thus enabling contact-state-aware tactile language understanding for contact-rich manipulation tasks. We first collect a novel dataset of 50k+ tactile 3D point cloud-language pairs, where descriptions explicitly capture multidimensional contact states (e.g., contact location, shape, and force) from the tactile sensor's perspective. CLTP leverages a pre-aligned and frozen vision-language feature space to bridge holistic textual and tactile modalities. Experiments validate its superiority in three downstream tasks: zero-shot 3D classification, contact state classification, and tactile 3D large language model (LLM) interaction. To the best of our knowledge, this is the first study to align tactile and language representations from the contact state perspective for manipulation tasks, providing great potential for tactile-language-action model learning. Code and datasets are open-sourced at https://sites.google.com/view/cltp/.

TLA: Tactile-Language-Action Model for Contact-Rich Manipulation

Significant progress has been made in vision-language models. However, language-conditioned robotic manipulation for contact-rich tasks remains underexplored, particularly in terms of tactile sensing. To address this gap, we introduce the Tactile-Language-Action (TLA) model, which effectively processes sequential tactile feedback via cross-modal language grounding to enable robust policy generation in contact-intensive scenarios. In addition, we construct a comprehensive dataset that contains 24k pairs of tactile action instruction data, customized for fingertip peg-in-hole assembly, providing essential resources for TLA training and evaluation. Our results show that TLA significantly outperforms traditional imitation learning methods (e.g., diffusion policy) in terms of effective action generation and action accuracy, while demonstrating strong generalization capabilities by achieving over 85\% success rate on previously unseen assembly clearances and peg shapes. We publicly release all data and code in the hope of advancing research in language-conditioned tactile manipulation skill learning. Project website: https://sites.google.com/view/tactile-language-action/

MapFusion: A Novel BEV Feature Fusion Network for Multi-modal Map Construction

Map construction task plays a vital role in providing precise and comprehensive static environmental information essential for autonomous driving systems. Primary sensors include cameras and LiDAR, with configurations varying between camera-only, LiDAR-only, or camera-LiDAR fusion, based on cost-performance considerations. While fusion-based methods typically perform best, existing approaches often neglect modality interaction and rely on simple fusion strategies, which suffer from the problems of misalignment and information loss. To address these issues, we propose MapFusion, a novel multi-modal Bird's-Eye View (BEV) feature fusion method for map construction. Specifically, to solve the semantic misalignment problem between camera and LiDAR BEV features, we introduce the Cross-modal Interaction Transform (CIT) module, enabling interaction between two BEV feature spaces and enhancing feature representation through a self-attention mechanism. Additionally, we propose an effective Dual Dynamic Fusion (DDF) module to adaptively select valuable information from different modalities, which can take full advantage of the inherent information between different modalities. Moreover, MapFusion is designed to be simple and plug-and-play, easily integrated into existing pipelines. We evaluate MapFusion on two map construction tasks, including High-definition (HD) map and BEV map segmentation, to show its versatility and effectiveness. Compared with the state-of-the-art methods, MapFusion achieves 3.6% and 6.2% absolute improvements on the HD map construction and BEV map segmentation tasks on the nuScenes dataset, respectively, demonstrating the superiority of our approach.

BCTR: Bidirectional Conditioning Transformer for Scene Graph Generation

Scene Graph Generation (SGG) remains a challenging task due to its compositional property. Previous approaches improve prediction efficiency by learning in an end-to-end manner. However, these methods exhibit limited performance as they assume unidirectional conditioning between entities and predicates, leading to insufficient information interaction. To address this limitation, we propose a novel bidirectional conditioning factorization for SGG, introducing efficient interaction between entities and predicates. Specifically, we develop an end-to-end scene graph generation model, Bidirectional Conditioning Transformer (BCTR), to implement our factorization. BCTR consists of two key modules. First, the Bidirectional Conditioning Generator (BCG) facilitates multi-stage interactive feature augmentation between entities and predicates, enabling mutual benefits between the two predictions. Second, Random Feature Alignment (RFA) regularizes the feature space by distilling multi-modal knowledge from pre-trained models, enhancing BCTR's ability on tailed categories without relying on statistical priors. We conduct a series of experiments on Visual Genome and Open Image V6, demonstrating that BCTR achieves state-of-the-art performance on both benchmarks. The code will be available upon acceptance of the paper.

HyCIR: Boosting Zero-Shot Composed Image Retrieval with Synthetic Labels

Composed Image Retrieval (CIR) aims to retrieve images based on a query image with text. Current Zero-Shot CIR (ZS-CIR) methods try to solve CIR tasks without using expensive triplet-labeled training datasets. However, the gap between ZS-CIR and triplet-supervised CIR is still large. In this work, we propose Hybrid CIR (HyCIR), which uses synthetic labels to boost the performance of ZS-CIR. A new label Synthesis pipeline for CIR (SynCir) is proposed, in which only unlabeled images are required. First, image pairs are extracted based on visual similarity. Second, query text is generated for each image pair based on vision-language model and LLM. Third, the data is further filtered in language space based on semantic similarity. To improve ZS-CIR performance, we propose a hybrid training strategy to work with both ZS-CIR supervision and synthetic CIR triplets. Two kinds of contrastive learning are adopted. One is to use large-scale unlabeled image dataset to learn an image-to-text mapping with good generalization. The other is to use synthetic CIR triplets to learn a better mapping for CIR tasks. Our approach achieves SOTA zero-shot performance on the common CIR benchmarks: CIRR and CIRCO.