News
Follow our latest publications, awards, field activities, student achievements, and research updates from IntelliSense Lab.
ACM MM 2026 ACM MM 2026 Oral | HiRS-Agent: A Hierarchical Multi-Agent System for Reliable Long-Horizon Remote Sensing Task Solving
HiRS-Agent puts a Manager over three Specialists so a long remote sensing workflow is planned, routed, executed and checked step by step. Verification at every step lifts long-horizon task accuracy from 15.73% to 43.95%. Accepted as an Oral at ACM Multimedia 2026.
ACM MM 2026 ACM MM 2026 | GeoChrono: Benchmarking and Rethinking Long-Term Temporal Understanding in Remote Sensing
ChronoBench breaks long-term remote sensing understanding into four cognitive levels and 12 sub-tasks, pinpointing long-term memory as the key bottleneck for MLLMs. GeoChrono reaches 78.34% overall accuracy, more than 20 points above the best commercial model.
ACM MM 2026 ACM MM 2026 | Self in Space: Benchmarking Self-Awareness and Spatial Cognition in UAV Embodied Intelligence
SIS-Bench evaluates spatial cognition and self-awareness of multimodal LLMs for UAV embodied intelligence across 1,646 real UAV videos and 4,856 QA pairs, while SIS-Motion probes whether explicit optical-flow motion cues help.
SCIS SCIS | RS-Agent: Automating Remote Sensing Tasks through Intelligent Agent
RS-Agent turns a large language model from a passive image-understanding tool into an agent that plans and executes remote sensing tasks, reaching over 95% planning accuracy across 9 datasets and 18 task types. Accepted by Science China Information Sciences.
Explainer Explainer | From Prompt to Policy: How Reinforcement Learning Teaches LLMs to Call Tools
A walk through the 2025 work on reinforcement learning for tool-integrated reasoning, from single-tool agents such as ReTool to multi-tool orchestration — and why prompting and supervised fine-tuning alone do not produce a reliable calling strategy.
Explainer Explainer | From Pre-training to Supervised Fine-tuning: How a Large Language Model Grows Up
The two foundations under every modern LLM, unpacked: what pre-training buys on a trillion-token corpus, how BERT and GPT split on architecture and objective, and what supervised fine-tuning adds on top.
Explainer Explainer | Near Real-Time Remote Sensing: A Space-Based Information Service for the Mega-Constellation Era
Classic remote sensing pipelines are too slow for disaster response: revisit times run to days, tasking needs people in the loop, and downlink bandwidth is scarce. This piece sets out a mega-constellation architecture — laser inter-satellite links, integrated communication and sensing payloads, and on-orbit compute — that closes the loop from request to product in 15 to 60 minutes.
No news match your search yet.