Safety
Hallucination-aware Faithful Reasoning in Remote Sensing
arXiv:2610.18369v1 Announce Type: new Abstract: Multimodal large language models (MLLMs) have demonstrated promising capabilities for remote sensing reasoning, yet they frequently generate visually un
arXiv:2610.18369v1 Announce Type: new Abstract: Multimodal large language models (MLLMs) have demonstrated promising capabilities for remote sensing reasoning, yet they frequently generate visually unsupported reasoning that contradicts the visual evidence. Existing post-training methods optimize primarily for final-answer correctness, providing limited supervision for identifying and correcting hallucinations that arise during the reasoning process. To address this limitation, we propose extbf{FaithRS}, a hallucination-aware self-reflection framework for faithful reasoning in remote sensing. We first construct extbf{SFHD-43K}, a dataset of 43K structured, fine-grained hallucination annotations, and train a detector to identify hallucination types, underlying causes, and supporting visual evidence. FaithRS then incorporates this diagnostic feedback into a Safe Detector-Guided Corrective Rollout framework, where a safety-aware gate selectively determines whether and how to revise the response. Training is driven by a composite reward that jointly optimizes answer accuracy, reasoning faithfulness, hallucination reduction, format compliance, and revision safety. Experiments with Qwen3.5 backbones (0.8B, 4B, and 9B) show consistent improvements across four remote sensing benchmarks, with gains of up to 11.64 points.
Related
- Mitigating Visual Hallucinations in Multimodal Systems through Retrieval-Augmented Reliability-Aware Inference
- SAVOR: Self-Aware Visual Grounding via Confidence-Calibrated Reinforcement Learning for Multimodal Hallucination Mitigation
- See Fair, Speak Truth: Equitable Attention Improves Grounding and Reduces Hallucination in Vision-Language Alignment
Source: arXiv cs.CV | 2026-10-14