EgoSafe-Bench is introduced, a benchmark specifically designed to probe forensic reasoning in egocentric safety scenarios, generated by pairing each of the 3,000 video clips with a QA chain governed by the proposed Hierarchical Reasoning Evaluation (HRE) protocol.
Abstract
Reliable visual safety understanding in real-world scenarios demands more than just object recognition; it requires causal reasoning under epistemic uncertainty. While Large Vision-Language Models (LVLMs) demonstrate impressive semantic alignment on standard benchmarks, they often struggle to distinguish between superficial correlation and genuine forensic logic when grounded in the dynamic, partially observable nature of first-person experiences. Existing evaluations, dominated by third-person surveillance footage and binary classification metrics, fail to expose this cognitive gap. To address this, we introduce EgoSafe-Bench, a benchmark specifically designed to probe forensic reasoning in egocentric safety scenarios. It comprises 12,000 unique evaluation samples, generated by pairing each of the 3,000 video clips with a QA chain governed by our proposed Hierarchical Reasoning Evaluation (HRE) protocol. Unlike standard benchmarks, HRE mandates a rigorous reasoning trajectory from initial feature anchoring to blind-spot deduction and intent inference, thereby enforcing logical consistency and penalizing shortcut-based predictions. Extensive evaluations of state-of-the-art LVLMs (e.g., Qwen3-VL, Gemini, VideoLLaMA 3) reveal a significant perception-reasoning decoupling: models often achieve high descriptive scores but exhibit notable fragility in causal reasoning and logical closure. Our work provides both a challenging dataset and a systematic evaluation framework to foster the development of logically robust video understanding systems.
CAViAR exposes a practical Perception--Reasoning Gap: current VLMs may recognize salient context, but do not reliably map visible agent actions to annotated rule-relevant responsibility categories in safety-critical driving scenarios, and all models degrade sharply on accident type and responsibility reasoning.
Sparsh Garg, Yi-Wen Chen, Vijay Kumar et al.· 0 citations
Accurate and trustworthy pedestrian detection is a foundational requirement for safe autonomous systems, particularly as environmental conditions become increasingly complex. However, conventional detection systems frequently struggle to maintain reliability under low-visibility conditions and are highly susceptible to false alarms caused by human-like interference, limiting their real-world trustworthiness. To overcome these perceptual bottlenecks, we propose STFPD, a novel interpretable semantic-target fusion strategy via vision-language self-supervision designed for multispectral pedestrian detection. Our framework introduces a tri-modal feature fusion methodology that integrates complementary evidence from RGB images, thermal signals, and text. Specifically, we utilize a self-supervised vision-language model to generate textual representations, explicitly modeling semantic context without requiring manual annotations. During the fusion phase, we exploit both parallel and cross-channel similarities among the three modalities, extracting effective representations through dynamic spatial sampling. Crucially, to ensure explainability and verifiable reasoning, we introduce a mask generation sub-network in the refinement phase, which enhances feature contrast and provides precise evidence localization. Extensive evaluations demonstrate that STFPD achieves outstanding performance and robust reliability. On the KAIST benchmark, STFPD achieves a competitive all-day miss rate of 4.45%, outperforming the Faster R-CNN baseline by 14.41% and the previous best model by 1.24%, while recording an exceptional night-time miss rate of only 3.33%. Furthermore, on the LLVIP dataset, it attains an Average Precision (AP) of 72.4% and an AP50 of 98.1%, yielding a substantial 6.1% AP improvement over existing methods. Qualitative 3D saliency visualizations further confirm that STFPD provides high-contrast target localization with near-zero false alarms, suggesting promising potential for deployment in safety-oriented intelligent transportation systems.
Recent visual generators produce high-fidelity images yet often violate physical consistency under ego-motion, limiting their use for spatial reasoning and embodied planning. Existing benchmarks largely focus on isolated images or single-step quality, leaving this challenge underexplored. We introduce EgoGenEval, a geometry-grounded, pose-free benchmark designed to evaluate the physical consistency of visual generators under ego-motion, and organize our study into two parts. (1) EgoGenEval contains 1,400 cases and 2,360 target views spanning single-step and multi-step ego-motion. It separately measures Camera Motion Grounding (CMG) and Scene State Preservation (SSP), with both metrics validated against blinded human judgments. Evaluating 16 pose-free generators together with two pose-conditioned references reveals that current models struggle to execute camera motion while maintaining scene state, and that no system performs well on both axes at once. (2) To examine whether benchmark-derived data can improve these capabilities, we build EgoGen-Train from the same geometry-grounded pipeline and run controlled SFT studies. These show that pairwise supervision does not reliably improve camera-motion grounding and scene-state preservation together: even at the full training pool and the longest budget, scene preservation gains a fraction of what camera motion does. This points to the pairwise teacher-forced objective itself as the binding constraint, motivating a trajectory-centric paradigm that couples self-conditioned rollouts with explicit pose and visibility supervision.
Yilin Long, Chenming Zhu, Zitang Gou et al.· 0 citations
Electronic travel aids are pivotal for the independent mobility of the visually impaired. While Vision-Language Models (VLMs) offer rich environmental understanding, they often suffer from excessive false positives in dynamic scenarios, leading to cognitive overload. To address this, we present ForeSightGuide, an anticipatory assistive guidance framework that couples semantic scene understanding with predictive hazard assessment. Unlike reactive systems, ForeSightGuide leverages the reasoning capabilities of VLMs to anticipate obstacle motion, effectively filtering out non-threatening objects to provide concise, actionable guidance. To validate our approach, we introduce a novel dataset captured in complex, dynamic real-world traffic scenes, designed to benchmark predictive capabilities. Extensive experiments on both public benchmarks and our proposed dataset demonstrate that ForeSightGuide achieves state-of-the-art performance. Notably, it significantly mitigates information overload by reducing redundant alerts to 0.299 per guidance output while maintaining a low missed-hazard rate of 0.112, proving its efficacy for safe walking assistance.
Zhiyuan Wang, Xu Li, Shikang Guo et al.· 0 citations
Vision-language models (VLMs) increasingly rely on point coordinates as a compact and executable interface for visual grounding in GUI interaction, robotic manipulation, and interactive visual systems. However, learning reliable pointing behavior remains difficult because the supervision space is inherently non-unique: many coordinates may be valid within the same target region, while multi-instance instructions require target coverage, count consistency, and duplicate suppression. This work presents PointRL, a verifiable reinforcement learning framework that learns point-level grounding from existing heterogeneous annotation evidence. PointRL converts bounding boxes, masks, and instance labels into pointing instructions, while retaining their target supports, instance membership, and set constraints as hidden verifier evidence, i.e., annotations kept outside the prompt and used by a deterministic checker to score predictions. The proposed reward evaluates parseability, point validity, instance coverage, cardinality consistency, and redundant or missing predictions. On PointArena, PointRL improves the overall accuracy of Qwen3.5-4B from 56.11% to 65.58%. Further evaluations on RoboSpatial, BLINK, and Ref-Adv show same-backbone gains on the evaluated external benchmarks, suggesting that verifiable point-level feedback may benefit spatial grounding in these settings.
Jing-Yang Su, Pu Cao, Xiu-Ze Jin et al.· 0 citations
UAV video question answering requires separating camera motion from changes in the scene, but RGB-only multimodal models receive no explicit, stable reference for that separation. We present EgoSIS, a pose-free adapter that converts RGB-derived bidirectional flow into motion-canonical visual evidence in three stages. Factorized Visual Ego-Transitions (FVET) fits a robust image-plane transition and exposes motion, residual-support, and reliability factors. Reliability-Gated Ego-Transition Memory (ReTEM) uses reliability-weighted updates for a bounded history and re-anchors it at cuts or sustained uncertainty. Ego-Aligned Spatial Evidence (EASE) warps supported visual features into each segment's local anchor and injects four spatial evidence tokens per visual slice through zero-initialized residuals, without changing Qwen's visual-token count. On SIS-Bench, EgoSIS-8B obtains 89.9\% perception, 82.5\% perception-plus-memory, and 76.2\% overall accuracy, with the largest gains concentrated in self-awareness perception and memory. The adapter thus provides an interpretable interface between optical flow and spatial reasoning.
Unknown authors· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
