端到端

arXiv 2603.21831v1

cs.RO自动驾驶端到端

Can a Robot Walk the Robotic Dog: Triple-Zero Collaborative Navigation for Heterogeneous Multi-Agent Systems

This paper presents Triple Zero Path Planning (TZPP), a collaborative navigation framework for heterogeneous multi-robot systems achieving zero training, zero prior knowledge, and zero simulation requirements. The system employs a coordinator-explorer architecture where a Unitree G1 humanoid robot performs task coordination while a Unitree Go2 quadruped robot explores and identifies feasible paths using guidance from a multimodal large language model. The framework is evaluated across diverse indoor and outdoor environments including obstacle-rich and landmark-sparse settings, demonstrating robust and human-comparable navigation efficiency. By eliminating reliance on traditional training and simulation pipelines, TZPP provides a practical approach for real-world deployment of heterogeneous robot cooperation with strong adaptability to unseen scenarios.

TZPP Authors

arXiv 2603.21723v1

cs.CL自动驾驶端到端

MIND: Multi-agent Inference for Negotiation Dialogue in Travel Planning

This paper introduces MIND (Multi-agent Inference for Negotiation Dialogue), a framework for simulating realistic consensus-building among travelers with heterogeneous preferences in travel planning scenarios. Grounded in Theory of Mind (ToM), MIND incorporates a Strategic Appraisal phase that achieves 90.2% accuracy in inferring opponent willingness from linguistic nuances. The framework demonstrates significant improvements over traditional Multi-Agent Debate (MAD) approaches, with a 20.5% improvement in High-w Hit and 30.7% increase in Debate Hit-Rate. Qualitative evaluations using LLM-as-a-Judge confirm superior performance in Rationality (68.8%) and Fluency (72.4%), achieving an overall win rate of 68.3%. This work effectively models human negotiation dynamics through advanced language understanding and multi-agent reasoning.

MIND Authors

arXiv 2603.21696v1

cs.CV自动驾驶端到端

Cross-Scenario Deraining Adaptation with Unpaired Data: Superpixel Structural Priors and Multi-Stage Pseudo-Rain Synthesis

Image deraining is a critical low-level computer vision task essential for robust outdoor surveillance and autonomous driving systems. While deep learning methods have shown success in aligned training settings, they typically experience significant performance degradation when applied to unseen Out-of-Distribution scenarios due to domain discrepancies between synthetic training data and real-world rain dynamics. This paper proposes a cross-scenario deraining adaptation framework that eliminates the need for paired rainy observations in target domains, utilizing only rain-free background images. The method incorporates a Superpixel Generation module that extracts stable structural priors from source domains using Simple Linear Iterative Clustering, enabling effective rain removal across diverse scenarios.

Anonymous

arXiv 2603.21661v1

cs.RO自动驾驶端到端

RTD-RAX: Fast, Safe Trajectory Planning for Systems under Unknown Disturbances

This paper presents RTD-RAX, a runtime-assurance extension of Reachability-based Trajectory Design that addresses two critical limitations of standard RTD implementations: conservatism from worst-case reachable-set overapproximations and inability to handle real-time disturbances. The framework combines a non-conservative RTD formulation for rapid goal-directed trajectory generation with mixed monotone reachability for fast disturbance-aware safety certification. When proposed trajectories fail safety verification under uncertainty, a repair procedure finds nearby safe trajectories that maintain progress toward the goal while guaranteeing safety under real-time disturbances. This approach enables provably safe, real-time trajectory planning for autonomous systems operating in uncertain environments.

S. K. S. Mitraz +3

arXiv 2603.21635v1

cs.CV自动驾驶端到端

Spatio-Temporal Attention Enhanced Multi-Agent DRL for UAV-Assisted Wireless Networks with Limited Communications

This paper addresses data transmission challenges in UAV-assisted wireless networks where multiple unmanned aerial vehicles serve as relays between ground users and a remote base station. The proposed delay-tolerant multi-agent deep reinforcement learning algorithm jointly optimizes trajectory planning, network formation, and transmission control while incorporating a delay-penalized reward mechanism to encourage inter-UAV information sharing. To handle information loss from unreliable channel conditions, a spatio-temporal attention mechanism predicts and recovers missing network state information, enhancing each UAV's situational awareness for improved collaboration and overall network throughput.

arXiv 2603.21594v1

cs.CV自动驾驶端到端

HACMatch Semi-Supervised Rotation Regression with Hardness-Aware Curriculum Pseudo Labeling

Regressing 3D rotations of objects from 2D images is a crucial yet challenging task, with broad applications in autonomous driving, virtual reality, and robotic control. This paper proposes a hardness-aware curriculum learning framework for semi-supervised rotation regression that dynamically selects pseudo-labeled samples based on their difficulty, progressing from easy to complex examples. The approach addresses limitations of rigid entropy-based pseudo-label filtering by introducing both multi-stage and adaptive curriculum strategies to effectively distinguish between reliable and unreliable unlabeled samples.

HACMatch Authors

arXiv 2603.21583v1

cs.CL自动驾驶端到端

A transformer architecture alteration to incentivise externalised reasoning

We propose a novel architectural modification and post-training pipeline for enhancing large language model reasoning capabilities by teaching models to truncate forward passes early. Our approach augments the standard transformer architecture with an early-exit mechanism at intermediate layers, enabling the model to exit at shallower layers when tokens can be predicted without deep computation. Through a calibration stage followed by reinforcement learning, we incentivize the model to exit as early as possible while preserving task performance. Preliminary experiments on small reasoning models demonstrate adaptive computation reduction across tokens, suggesting that at appropriate scale, this approach can minimize excess computation for non-myopic planning using internal activations, reserving deep computation only for difficult-to-predict tokens.

Anonymous

arXiv 2603.21376v1

cs.RO端到端Transformer

Evaluating Factor-Wise Auxiliary Dynamics Supervision for Latent Structure and Robustness in Simulated Humanoid Locomotion

This paper evaluates whether factor-wise auxiliary dynamics supervision produces useful latent structure or improved robustness in simulated humanoid locomotion. We introduce DynaMITE, a transformer encoder with a factored 24-dimensional latent space trained using per-factor auxiliary losses during proximal policy optimization. Our method is compared against LSTM, plain Transformer, and MLP baselines on a Unitree G1 humanoid robot across four Isaac Lab tasks. Through comprehensive ablation studies with 10 random seeds, we analyze the contributions of tanh bottlenecks and auxiliary losses to in-distribution reward performance. Results demonstrate that the supervised latent fails to produce decodable or functionally separable factor structure, with probe R-squared near zero and minimal reward changes when subspaces are clamped.

arXiv 2603.21268v1

cs.CV自动驾驶端到端

Enhancing Brain Tumor Classification Using Vision Transformers with Colormap-Based Feature Representation on BRISC2025 Dataset

This study presents a deep learning framework for multi-class brain tumor classification from magnetic resonance imaging scans using Vision Transformers enhanced with colormap-based feature representation. The proposed method leverages transformer architectures to capture long-range dependencies while incorporating color mapping techniques to emphasize structural and intensity variations in MRI scans. Experiments on the BRISC2025 dataset containing glioma, meningioma, pituitary tumor, and non-tumor cases demonstrate superior performance with 98.90% accuracy, significantly outperforming baseline approaches. The framework utilizes standard metrics including accuracy, precision, recall, F1-score, and AUC for comprehensive evaluation.

arXiv 2603.21234v1

cs.CV端到端CV

GAPG: Geometry Aware Push-Grasping Synergy for Goal-Oriented Manipulation in Clutter

Grasping target objects is a fundamental skill for robotic manipulation, but in cluttered environments with stacked or occluded objects, a single-step grasp is often insufficient. This paper proposes a geometry-aware push-grasp synergy framework that leverages point cloud data to integrate grasp and push evaluation. The grasp evaluation module analyzes the geometric relationship between the gripper's point cloud and the points enclosed within its closing region to determine grasp feasibility and stability. Guided by this analysis, the push evaluation module predicts how pushing actions influence future graspable space, enabling the robot to effectively manipulate objects in complex cluttered scenarios.

GAPG Authors

arXiv 2603.21195v1

cs.CV自动驾驶端到端

Single-Eye View: Monocular Real-time Perception Package for Autonomous Driving

This paper presents LRHPerception, a real-time monocular perception package designed specifically for autonomous driving applications. The system leverages monocular camera video to interpret the surrounding environment by combining the computational efficiency of end-to-end learning with detailed local mapping methodologies. It achieves comprehensive perception through integrated object tracking and prediction, road segmentation, and depth estimation capabilities. The unified framework processes monocular images into a five-channel tensor containing RGB, road segmentation, and pixel-level depth information, enhanced with object detection and trajectory prediction. Experimental results demonstrate impressive real-time performance at 29 FPS on a single GPU, representing a 555% speedup compared to baseline approaches.

LRHPerception Team

arXiv 2603.21061v1

cs.CL大语言模型端到端

Mitigating Selection Bias in Large Language Models via Permutation-Aware GRPO

Large language models often exhibit selection bias in multiple-choice and pairwise evaluation tasks due to non-semantic factors such as option positions and label symbols. Existing inference-time debiasing methods are computationally expensive and may harm reasoning capabilities, while pointwise training approaches ignore the importance of consistent answers across different question permutations. This paper proposes Permutation-Aware Group Relative Policy Optimization (PA-GRPO), which addresses selection bias by enforcing permutation-consistent semantic reasoning through two complementary mechanisms: cross-permutation advantage computation relative to the mean reward over all permutations, and consistency-aware reward that encourages stable decision-making across different permutations. The proposed approach effectively mitigates selection bias while preserving the model's reasoning capabilities.

PA-GRPO Authors

arXiv 2603.21016v1

cs.AI端到端Transformer

Detection of adversarial intent in Human-AI teams using LLMs

Large language models (LLMs) are increasingly deployed in human-AI teams as support agents for complex tasks such as information retrieval, programming, and decision-making assistance. This paper studies the potential role of LLMs as defensive supervisors within mixed human-AI teams to detect malicious behavior. Using a dataset consisting of multi-party conversations and decisions over a 25-round horizon, we formulate the problem of malicious behavior detection from interaction traces. We find that LLMs are capable of identifying malicious behavior effectively, demonstrating their potential as defensive actors in collaborative environments.

Unknown

arXiv 2603.20976v1

cs.CL端到端Transformer

Do LLM-Driven Agents Exhibit Engagement Mechanisms? Controlled Tests of Information Load, Descriptive Norms, and Popularity Cues

Large language models enable increasingly expressive agent-based simulations, but pose methodological challenges regarding behavioral validity. This paper evaluates LLM-driven simulation credibility through a social media test case examining information engagement. Using a Weibo-like environment, the study systematically manipulates information load and descriptive norms while allowing popularity cues to evolve endogenously. The research tests whether simulated user behavior responds systematically to theoretical constructs rather than producing merely plausible outputs. Findings indicate that engagement responds systematically to information load and descriptive norms, with sensitivity to popularity cues varying across contexts. The paper discusses methodological implications for simulation-based communication research, particularly for multi-condition experimental designs involving LLM-driven agents.

arXiv 2603.20911v1

cs.CL大语言模型端到端

RubricRAG: Towards Interpretable and Reliable LLM Evaluation via Domain Knowledge Retrieval for Rubric Generation

Large language models are increasingly evaluated using automated graders that output scalar scores or preferences, but these approaches lack interpretability as a single score cannot explain why an answer is good or bad. Rubric-based evaluation offers a more transparent alternative by decomposing quality into explicit, checkable criteria. However, manually designing high-quality, query-specific rubrics is labor-intensive and cognitively demanding. This work investigates whether LLMs can generate interpretable and effective rubrics through domain knowledge retrieval for automated evaluation. The proposed RubricRAG framework aims to enhance the interpretability and reliability of LLM evaluation by leveraging retrieved domain knowledge to generate task-specific evaluation rubrics.

RubricRAG Authors

arXiv 2603.20882v1

cs.CV端到端CV

Speedup Patch: Learning a Plug-and-Play Policy to Accelerate Embodied Manipulation

This paper presents Speedup Patch (SuP), a lightweight policy-agnostic framework designed to accelerate embodied manipulation by adaptively downsampling action chunks from existing policies. The method formulates the scheduler optimization as a Constrained Markov Decision Process to maximize efficiency while maintaining task performance. To address offline safety constraints, the approach introduces World Model based state deviation as a surrogate metric for success evaluation. SuP demonstrates that embodied manipulation tasks can be significantly accelerated without requiring policy retraining or costly online interactions.

SuP Authors