An official website of the United States government
Large Vision-Language Models (LVLMs) have emerged as transformative tools in multimodal tasks, seamlessly integrating pretrained vision encoders to align visual and textual modalities. Prior works have highlighted the susceptibility of LVLMs to dual exploits (gradient-based and optimization-based jailbreak attacks), which leverage the expanded attack surface introduced by the image modality. Despite advancements in enhancing robustness, existing methods fall short in their ability to defend against dual exploits while preserving fine-grained semantic details and overall semantic coherence under intense adversarial perturbations. To bridge this gap, we introduce SheiLding against Dual Exploits (SLADE), a novel unsupervised adversarial fine-tuning scheme that enhances the resilience of CLIP-based vision encoders. SLADE’s dual-level contrastive learning approach balances the granular and the holistic, capturing fine-grained image details without losing sight of high-level semantic coherence. Extensive experiments demonstrate that SLADE-equipped LVLMs set a new benchmark for robustness against dual exploits while preserving fine-grained semantic details of perturbed images. Notably, SLADE achieves these results without compromising the core functionalities of LVLMs, such as instruction following, or requiring the computational overhead (e.g., large batch sizes, momentum encoders) commonly associated with traditional contrastive learning methods.
more »
« less
This content will become publicly available on June 1, 2027
Flat-Pack Bench: Evaluating Spatio-Temporal Understanding in Large Vision-Language Models through Furniture Assembly
The emergence of Large Vision-Language Models (LVLMs) has significantly advanced video understanding capabilities. However, existing benchmarks focus predominantly on coarse-grained tasks such as action segmentation, classification, captioning, and retrieval. Furthermore, these benchmarks often rely on entities that can be easily identified verbally, like household objects, animals, human subjects, etc., limiting their applicability to complex, in-the-wild video scenarios. But, many applications such as furniture assembly, cooking, etc., require step-by-step fine-grained spatio-temporal understanding of the video, which is not sufficiently evaluated in current benchmarks.To address this gap, we introduce Flat-Pack Bench, a novel benchmark centered on furniture assembly tasks. Our benchmark evaluates LVLMs on nuanced tasks, including temporal ordering of assembly actions, temporal localization of assembly state, understanding part mating, and tracking, using multiple-choice questions paired with visual prompts highlighting relevant parts as references for fine-grained questions. Our experiments reveal that state-of-the-art LVLMs struggle significantly with fine-grained spatio-temporal reasoning, highlighting their limitations in effectively leveraging temporal information from videos, limited tracking ability, and understanding of spatial interactions like physical contact.
more »
« less
- PAR ID:
- 10704343
- Publisher / Repository:
- Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
- Date Published:
- Page Range / eLocation ID:
- 16624-16634
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Understanding and reasoning about code semantics is essential for enhancing code LLMs' abilities to solve real-world software engineering (SE) tasks. Although several code reasoning benchmarks exist, most rely on synthetic datasets or educational coding problems and focus on coarse-grained reasoning tasks such as input/output prediction, limiting their effectiveness in evaluating LLMs in practical SE contexts. To bridge this gap, we propose CodeSense, the first benchmark that makes available a spectrum of fine-grained code reasoning tasks concerned with the software engineering of real-world code. We collected Python, C and Java software projects from real-world repositories. We executed tests from these repositories, collected their execution traces, and constructed a ground truth dataset for fine-grained semantic reasoning tasks. We then performed comprehensive evaluations on state-of-the-art LLMs. Our results show a clear performance gap for the models to handle fine-grained reasoning tasks. Although prompting techniques such as chain-of-thought and in-context learning helped, the lack of code semantics in LLMs fundamentally limit models' capabilities of code reasoning. Besides dataset, benchmark and evaluation, our work produced an execution tracing framework and tool set that make it easy to collect ground truth for fine-grained SE reasoning tasks, offering a strong basis for future benchmark construction and model post training. Our code and data are located at \url{https://codesense-bench.github.io/}.more » « less
-
While image captioning provides isolated descriptions for individual images, and video captioning offers one single narrative for an entire video clip, our work explores an important middle ground: progress-aware video captioning at the frame level. This novel task aims to generate temporally fine-grained captions that not only accurately describe each frame but also capture the subtle progression of actions throughout a video sequence. Despite the strong capabilities of existing leading vision language models, they often struggle to discern the nuances of frame-wise differences. To address this, we propose ProgressCaptioner, a captioning model designed to capture the fine-grained temporal dynamics within an action sequence. Alongside, we develop the FrameCap dataset to support training and the FrameCapEval benchmark to assess caption quality. The results demonstrate that ProgressCaptioner significantly surpasses leading captioning models, producing precise captions that accurately capture action progression and set a new standard for temporal precision in video captioning. Finally, we showcase practical applications of our approach, specifically in aiding keyframe selection and advancing video understanding, highlighting its broad utility.more » « less
-
Efficiently re-identifying and tracking objects across a network of cameras is crucial for applications like traffic surveillance. Spatula is the state-of-the-art video database man- agement system (VDBMS) for processing RE-ID queries. However, it suffers from two limitations. Its spatio-temporal filtering scheme has limited accuracy on large camera networks due to localized camera history. It is not suitable for critical video analytics applications that require high recall due to lack of support for adaptive query processing. In this paper, we present TRACER, a novel VDBMS for efficiently processing RE-ID queries using an adaptive query pro- cessing framework. TRACER selects the optimal camera to process at each time step by training a recurrent network to model long- term historical correlations. To accelerate queries under a high recall constraint, TRACER incorporates a probabilistic adaptive search model that processes camera feeds in incremental search windows and dynamically updates the sampling probabilities using an exploration-exploitation strategy. To address the paucity of benchmarks for the RE-ID task due to privacy concerns, we present a novel synthetic benchmark for generating multi-camera RE-ID datasets based on real-world traffic distribution. Our evaluation shows that TRACER outperforms the state-of-the-art cross-camera analytics system by 3.9⇥ on average across diverse datasets.more » « less
-
Fuzzing is a powerful software testing technique renowned for its effectiveness in identifying software vulnerabilities. Traditional fuzzing evaluations typically focus on overall fuzzer performance across a set of target programs, yet few benchmarks consider how fine-grained program features influence fuzzing effectiveness. To bridge this gap, we introduce FeatureBench, a novel benchmark designed to generate programs with configurable, fine-grained program features to enhance fuzzing evaluations. We reviewed 25 recent grey-box fuzzing studies, extracting 7 program features related to control-flow and data-flow that can impact fuzzer performance. Using these features, we generated a benchmark consisting of 153 programs controlled by 10 fine-grained configurable parameters. We evaluated 11 fuzzers using this benchmark, with each fuzzer representing either distinct claimed improvements or serving as a widely used baseline in fuzzing evaluations. The results indicate that fuzzer performance varies significantly based on the program features and their strengths, highlighting the importance of incorporating program characteristics into fuzzing evaluations.more » « less
