Public emergencies often trigger large-scale panic emotions, which are further amplified by social media, potentially leading to serious consequences such as resource hoarding and social disorder. Therefore, timely identification and intervention of panic are of great significance. However, the development of panic emotion recognition models is severely constrained by the scarcity of high-quality annotated datasets. Existing coarse-grained sentiment classification fails to distinguish panic from fear and anxiety, where the latter two are characterized by high arousal and a sense of loss of control. Furthermore, purely manual annotation is expensive and hard to scale, while fully automated annotation suffers from model hallucinations and conservative bias. To address the challenges in constructing large-scale panic emotion datasets, this paper leverages large language models (LLMs) and proposes a four-stage progressive human-AI collaborative annotation framework named HLLC (Human-LLM Collaborative Labeling). The framework sequentially performs data sampling, LLM initial labeling, crowdsourced verification, and model generalization. Through representative sampling, multi-model consistency comparison for zero-shot labeling, confidence-based sample routing, and BERT fine-tuning, the framework balances annotation quality and efficiency. Experiments on the CrisisNLP Hurricane Sandy dataset [1] show that the HLLC framework achieves a panic F1 score of 0.787 and an overall accuracy of 0.857, improving by 33.3% over lexicon methods and boosting panic recall by 78.7% over pure LLM annotation. These results demonstrate that HLLC enables large-scale, fine-grained panic emotion dataset construction at low cost, offering a practical solution for emergency public opinion monitoring and early warning.
Yu Cai, Chuan Ai, Meng-Zhu Liu et al.· 2026 12th International Conf...· 0 citations
Language-goal aerial navigation requires an agent to local- ize a potentially unobserved target from relational instruc- tions and partial observations, and translate this inference into metric actions in large-scale continuous environments. Existing methods often reduce language grounding to one single waypoint or action, prematurely collapsing the spatial uncertainty inherent in incomplete evidence and ambiguous relations. To address this limitation, we introduce SBFNav, a closed-loop navigation framework centered on a language- conditioned Spatial Belief Field (SBF). Unlike ego-centric maps that primarily record what has been observed, SBF rep- resents a task-conditioned distribution over plausible target locations, preserving multiple spatial hypotheses under par- tial evidence. At each step, this distribution is updated from accumulated observations as new evidence becomes avail- able. Built on this representation, SBFNav selects the goal that best aligns with the instruction and observations as a met- ric waypoint for control. Experiments on both the original and revised CityNav benchmarks achieve the best reported overall performance. On the Test Unseen split, our method improves SR from 25.91% to 32.29% and SPL from 19.63% to 30.43%. Ablation studies further confirm the advantages of spatial-belief modeling over single-point prediction.
Hao-Tian Xu, Yue Hu, Zheng-Qiu Zhu et al.· 0 citations
Air-Ground Object Search (AGOS) in urban environments is a challenging embodied task, which requires an Unmanned Aerial Vehicle (UAV) and an Unmanned Ground Vehicle (UGV) to jointly search for and verify a specified target vehicle from multi-view visual references. To study this underexplored problem, we introduce AGOS-Bench, the first dedicated benchmark for evaluating whether general-purpose Vision-Language Models (VLMs) can integrate aerial discoveries and ground-level verification through UAV-UGV cooperation. We further provide AGOS-Dataset as the companion resource of exemplary trajectories constructed by an automatic pipeline. It consists of 7.7k episodes for searching objects of diverse categories and attributes, spanning three difficulty levels. To address the AGOS task, we propose AGOS-Agent, a training-free and tool-augmented approach. The agentic method relieves VLMs from complex and dynamic coordination via a deliberate search-handoff-verify cooperation protocol, only demanding VLMs for scene understanding and decision-making. Extensive experiments on nine VLMs show that AGOS-Agent improves overall success rate for eight of the nine evaluated backbones while reducing decision steps for all nine. On the hard split, the SR and SPL of Gemini-3.6-Flash increase from 8.6% to 55.7% and from 7.6% to 44.0%, respectively.
