Space Embodied Intelligence

Guidance, navigation, control and thermal management in one system view.

AI for Space: A Survey of Guidance, Navigation, Control, and Thermal Management

Connect model architectures, physical adaptation and validation.

Foundation Models for Autonomous Space Robotics: Architectures, Physical Adaptation, and Validation

Make physical resource budgets part of every autonomous decision.

Resource-Aware Embodied Autonomy for Space Robots: A Review

Adjust model weights in the frequency domain without changing the inference architecture.

Unleashing the Potential of Large Language Models through Spectral Modulation

Analyze knowledge changes across model layers to identify critical and redundant computation.

Towards Understanding How Knowledge Evolves in Large Vision-Language Models

Route uncertain classification results to a diffusion model for further evaluation.

An Efficient Framework for Enhancing Discriminative Models via Diffusion Techniques

Compose one static low-bit model from AWQ, GPTQ and AutoRound checkpoints, accepting each module substitution only when it closes the KL gap to the BF16 reference.

Compressing Onboard Language Models for Space Robots under a Behavioral-Fidelity Requirement

Estimate how image patches contribute to a vision-language model’s answer.

Patchwise Cooperative Game-based Interpretability Method for Large Vision-language Models

Generate adversarial surface patterns to test object detector robustness.

AdvLogo: Adversarial Patch Attack against Object Detectors based on Diffusion Models

Evaluate reasoning quality and decision latency in a continuously changing environment.

Beyond Scaling: Assessing Strategic Reasoning and Rapid Decision-Making Capability of LLMs in Zero-sum Environments

Compare candidate responses through model peer review and calibrated scoring.

Scoring, Reasoning, and Selecting the Best! Ensembling Large Language Models via a Peer-Review Process

Paired-outcome learning that adapts maneuver speed to uncertain braking authority.

Paired-Outcome Learning for Selective Spacecraft Attitude Capture under Torque Limits

Learning waypoints and passage events jointly for cooperative local planning.

Learning Waypoints and Passage Events for Decentralized Multi-Quadrotor Trajectory Planning

Generate coupled states and actions, with explicit control over scenario difficulty.

End-to-end Conditional Diffusion for Realistic and Controllable Visual Traffic Scenario Generation

Optimize initial noise to improve agreement between generated images and text requirements.

Noise Diffusion for Enhancing Semantic Faithfulness in Text-to-Image Synthesis