
Conference
IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR)
Colorado Convention Center, Denver, Colorado
The IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR) is the premier annual computer vision event comprising the main conference and several co-located workshops and short courses. This year’s conference will be held at the Colorado Convention Center in Denver, Colorado from June 3–7, 2026.
IAI Researchers and their collaborators have 13 accepted papers in CVPR 2026.
The h5-index is the h-index for articles published in the last 5 complete years. According to Google Scholar Metrics, Computer Vision and Pattern Recognition (CVPR) is ranked 2nd in the h5-index rankings for top venues.
You can access the IAI Publications Page for enhanced search capabilities and view papers below.
Accepted Papers
Luo, Jinqi; Yang, Jinyu; Neiman, Tal; Fan, Lei; Yina, Bing; Tran, Son; Shah, Mubarak; Vidal, Rene
Dictionary-Aligned Concept Control for Safeguarding Multimodal LLMs Conference
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026), 2026.
@conference{nokey,
title = {Dictionary-Aligned Concept Control for Safeguarding Multimodal LLMs},
author = {Jinqi Luo and Jinyu Yang and Tal Neiman and Lei Fan and Bing Yina and Son Tran and Mubarak Shah and Rene Vidal},
url = {https://arxiv.org/abs/2604.08846},
year = {2026},
date = {2026-04-10},
urldate = {2026-04-10},
publisher = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026)},
abstract = {Multimodal Large Language Models (MLLMs) have been shown to be vulnerable to malicious queries that can elicit unsafe responses. Recent work uses prompt engineering, response classification, or finetuning to improve MLLM safety. Nevertheless, such approaches are often ineffective against evolving malicious patterns, may require rerunning the query, or demand heavy computational resources. Steering the activations of a frozen model at inference time has recently emerged as a flexible and effective solution. However, existing steering methods for MLLMs typically handle only a narrow set of safety-related concepts or struggle to adjust specific concepts without affecting others. To address these challenges, we introduce Dictionary-Aligned Concept Control (DACO), a framework that utilizes a curated concept dictionary and a Sparse Autoencoder (SAE) to provide granular control over MLLM activations. First, we curate a dictionary of 15,000 multimodal concepts by retrieving over 400,000 caption-image stimuli and summarizing their activations into concept directions. We name the dataset DACO-400K. Second, we show that the curated dictionary can be used to intervene activations via sparse coding. Third, we propose a new steering approach that uses our dictionary to initialize the training of an SAE and automatically annotate the semantics of the SAE atoms for safeguarding MLLMs. Experiments on multiple MLLMs (e.g., QwenVL, LLaVA, InternVL) across safety benchmarks (e.g., MM-SafetyBench, JailBreakV) show that DACO significantly improves MLLM safety while maintaining general-purpose capabilities.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Kulkarni, Parth Parag; Gupta, Rohit; Chhipa, Prakash Chandra; Shah, Mubarak
VidTAG: Temporally Aligned Video to GPS Geolocalization with Denoising Sequence Prediction at a Global Scale Conference
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026), 2026.
@conference{nokey,
title = {VidTAG: Temporally Aligned Video to GPS Geolocalization with Denoising Sequence Prediction at a Global Scale},
author = {Parth Parag Kulkarni and Rohit Gupta and Prakash Chandra Chhipa and Mubarak Shah},
url = {https://arxiv.org/abs/2604.12159},
year = {2026},
date = {2026-04-10},
urldate = {2026-04-10},
publisher = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026)},
abstract = {The task of video geolocalization aims to determine the precise GPS coordinates of a video’s origin and map its trajectory; with applications in forensics, social media, and exploration. Existing classification-based approaches operate at a coarse city-level granularity and fail to capture fine-grained details, while image retrieval methods are impractical on a global scale due to the need for extensive image galleries which are infeasible to compile. Comparatively, constructing a gallery of GPS coordinates is straightforward and inexpensive. We propose VidTAG, a dual-encoder framework that performs frameto-GPS retrieval using both self-supervised and languagealigned features. To address temporal inconsistencies in video predictions, we introduce the TempGeo module, which aligns frame embeddings, and the GeoRefiner module, an encoder-decoder architecture that refines GPS features using the aligned frame embeddings. Evaluations on Mapillary (MSLS) and GAMa datasets demonstrate our model’s ability to generate temporally consistent trajectories and outperform baselines, achieving a 20% improvement at the 1 km threshold over GeoCLIP. We also beat current Stateof-the-Art by 25% on global coarse grained video geolocalization (CityGuessr68k). Our approach enables finegrained video geolocalization and lays a strong foundation for future research. More details on the project webpage: parthpk.github.io/vidtag webpage.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Chan-Santiago, Jeffrey A.; Shah, Mubarak
Learnability-Guided Diffusion for Dataset Distillation Conference
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026), 2026.
@conference{nokey,
title = {Learnability-Guided Diffusion for Dataset Distillation},
author = {Jeffrey A. Chan-Santiago and Mubarak Shah},
url = {https://arxiv.org/abs/2604.00519},
year = {2026},
date = {2026-04-01},
publisher = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026)},
abstract = {Training machine learning models on massive datasets is expensive and time-consuming. Dataset distillation addresses this by creating a small synthetic dataset that achieves the same performance as the full dataset. Recent methods use diffusion models to generate distilled data, either by promoting diversity or matching training gradients. However, existing approaches produce redundant training signals, where samples convey overlapping information. Empirically, disjoint subsets of distilled datasets capture 80-90% overlapping signals. This redundancy stems from optimizing visual diversity or average training dynamics without accounting for similarity across samples, leading to datasets where multiple samples share similar information rather than complementary knowledge. We propose learnability-driven dataset distillation, which constructs synthetic datasets incrementally through successive stages. Starting from a small set, we train a model and generate new samples guided by learnability scores that identify what the current model can learn from, creating an adaptive curriculum. We introduce Learnability-Guided Diffusion (LGD), which balances training utility for the current model with validity under a reference model to generate curriculum-aligned samples. Our approach reduces redundancy by 39.1%, promotes specialization across training stages, and achieves state-of-the-art results on ImageNet-1K (60.1%), ImageNette (87.2%), and ImageWoof (72.9%). Our code is available on our project page this https URL.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Swetha, Sirnam; Gupta, Rohit; Kulkarni, Parth Parag; Shatwell, David G; Santiago, Jeffrey A Chan; Siddiqui, Nyle; Fioresi, Joseph; Shah, Mubarak
VRR-QA: Visual Relational Reasoning in Videos Beyond Explicit Cues Conference
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026), 2026.
@conference{nokey,
title = {VRR-QA: Visual Relational Reasoning in Videos Beyond Explicit Cues},
author = {Sirnam Swetha and Rohit Gupta and Parth Parag Kulkarni and David G Shatwell and Jeffrey A Chan Santiago and Nyle Siddiqui and Joseph Fioresi and Mubarak Shah},
url = {https://arxiv.org/abs/2506.21742},
year = {2026},
date = {2026-03-29},
publisher = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026)},
abstract = {Video Question Answering (VideoQA) has made significant strides by leveraging multimodal learning to align visual and textual modalities. However, current benchmarks overwhelmingly focus on questions answerable through explicit visual content - actions, objects, and events - directly observable within individual frames or short clips. To truly understand videos as humans do, models must go beyond what is directly shown, inferring hidden relationships and contextual cues that are only implied across frames. Current benchmarks fail to capture this essential aspect of video understanding. To address this gap, we introduce VRR-QA, a benchmark for Visual Relational Reasoning Beyond Explicit Cues. We curate our benchmark from creative and cinematic videos such as movies, that deliberately employ storytelling techniques which omit direct depictions of certain events or relations, requiring viewers to infer them. VRR-QA comprises 1K meticulously expert-annotated QA pairs drawn from 1K creative video clips covering 15 genres across 7 decades of content, from both live-action and animated titles. Our extensive evaluations on 14 leading VideoQA models reveals consistent and significant performance degradation, underscoring their reliance on surface-level visual cues and highlighting the difficulty of implicit reasoning. Even the best model substantially underperforms human baselines with only 64% accuracy. Performance variations across models further illustrate the complexity and diversity of the challenges presented by VRR-QA. By releasing both dataset and data collection framework, VRR-QA establishes a rigorous, diverse, and reproducible testbed for advancing VideoQA: this https URL.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Shatwell, David G.; Swetha, Sirnam; Shah, Mubarak
TIGeR: A Unified Framework for Time, Images and Geo-location Retrieval Conference
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026), 2026.
@conference{nokey,
title = {TIGeR: A Unified Framework for Time, Images and Geo-location Retrieval},
author = {David G. Shatwell and Sirnam Swetha and Mubarak Shah},
url = {https://arxiv.org/abs/2603.24749},
year = {2026},
date = {2026-03-28},
urldate = {2026-03-28},
publisher = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026)},
abstract = {Many real-world applications in digital forensics, urban monitoring, and environmental analysis require jointly reasoning about visual appearance, location, and time. Beyond standard geo-localization and time-of-capture prediction, these applications increasingly demand more complex capabilities, such as retrieving an image captured at the same location as a query image but at a specified target time. We formalize this problem as Geo-Time Aware Image Retrieval and propose TIGeR, a unified framework for Time, Images and Geo-location Retrieval. TIGeR supports flexible input configurations (single-modality and multi-modality queries) and uses the same representation to perform (i) geo-localization, (ii) time-of-capture prediction, and (iii) geo-time-aware retrieval. By preserving the underlying location identity despite large appearance changes, TIGeR enables retrieval based on where and when a scene was captured, rather than purely on visual similarity. To support this task, we design a multistage data curation pipeline and propose a new diverse dataset of 4.5M paired image-location-time triplets for training and 86k high-quality triplets for evaluation. Extensive experiments show that TIGeR consistently outperforms strong baselines and state-of-the-art methods by up to 16% on time-of-year, 8% time-of-day prediction, and 14% in geo-time aware retrieval recall, highlighting the benefits of unified geo-temporal modeling.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Luan, Siyu; Li, Yan; Chena, Zhong; Wang, Zhenyi
Adaptive Bayesian Early-Exit Networks for Efficient Non-Transferable Learning Conference
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026), 2026.
@conference{nokey,
title = {Adaptive Bayesian Early-Exit Networks for Efficient Non-Transferable Learning},
author = {Siyu Luan and Yan Li and Zhong Chena and Zhenyi Wang},
year = {2026},
date = {2026-03-26},
urldate = {2026-03-26},
publisher = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026)},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Liu, Yuyuan; Chen, Yuanhong; Wang, Chong; Han, Junlin; Wu, Junde; Peng, Can; Chen, Jingkun; Tian, Yu; Carneiro, Gustavo
AuralSAM2: Enabling SAM2 Hear Through Pyramid Audio-Visual Feature Prompting Conference
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026), 2026.
@conference{nokey,
title = {AuralSAM2: Enabling SAM2 Hear Through Pyramid Audio-Visual Feature Prompting},
author = {Yuyuan Liu and Yuanhong Chen and Chong Wang and Junlin Han and Junde Wu and Can Peng and Jingkun Chen and Yu Tian and Gustavo Carneiro},
url = {https://arxiv.org/abs/2506.01015},
year = {2026},
date = {2026-03-26},
urldate = {2026-03-26},
publisher = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026)},
abstract = {Segment Anything Model 2 (SAM2) exhibits strong generalisation for promptable segmentation in video clips; however, its integration with the audio modality remains underexplored. Existing approaches mainly follow two directions: (1) injecting adapters into the image encoder to receive audio signals, which incurs efficiency costs during prompt engineering, and (2) leveraging additional foundation models to generate visual prompts for the sounding objects, which are often imprecisely localised, leading to misguidance in SAM2. Moreover, these methods overlook the rich semantic interplay between hierarchical visual features and other modalities, resulting in suboptimal cross-modal fusion. In this work, we propose AuralSAM2, comprising the novel AuralFuser module, which externally attaches to SAM2 to integrate features from different modalities and generate feature-level prompts, guiding SAM2's decoder in segmenting sounding targets. Such integration is facilitated by a feature pyramid, further refining semantic understanding and enhancing object awareness in multimodal scenarios. Additionally, the audio-guided contrastive learning is introduced to explicitly align audio and visual representations and to also mitigate biases caused by dominant visual patterns. Results on public benchmarks show that our approach achieves remarkable improvements over the previous methods in the field. Code is available at this https URL.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Zhang, Yancheng; Zhang, Xiaohan; Sun, Guangyu; Lyu, Zonglin; Wshah, Safwan; Chen, Chen
Geo2: Geometry-Guided Cross-view Geo-Localization and Image Synthesis Conference
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026), 2026.
@conference{nokey,
title = {Geo2: Geometry-Guided Cross-view Geo-Localization and Image Synthesis},
author = {Yancheng Zhang and Xiaohan Zhang and Guangyu Sun and Zonglin Lyu and Safwan Wshah and Chen Chen},
url = {https://arxiv.org/html/2603.25819v1},
year = {2026},
date = {2026-03-26},
urldate = {2026-03-26},
publisher = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026)},
abstract = {Cross-view geo-spatial learning consists of two important tasks: Cross-View Geo-Localization (CVGL) and Cross-View Image Synthesis (CVIS), both of which rely on establishing geometric correspondences between ground and aerial views. Recent Geometric Foundation Models (GFMs) have demonstrated strong capabilities in extracting generalizable 3D geometric features from images, but their potential in cross-view geo-spatial tasks remains underexplored. In this work, we present Geo
2
, a unified framework that leverages Geometric priors from GFMs (e.g., VGGT) to jointly perform Geo-spatial tasks, CVGL and bidirectional CVIS. Despite the 3D reconstruction ability of GFMs, directly applying them to CVGL and CVIS remains challenging due to the large viewpoint gap between ground and aerial imagery. We propose GeoMap, which embeds ground and aerial features into a shared 3D-aware latent space, effectively reducing cross-view discrepancies for localization. This shared latent space naturally bridges cross-view image synthesis in both directions. To exploit this, we propose GeoFlow, a flow-matching model conditioned on geometry-aware latent embeddings. We further introduce a consistency loss to enforce latent alignment between the two synthesis directions, ensuring bidirectional coherence. Extensive experiments on standard benchmarks, including CVUSA, CVACT, and VIGOR, demonstrate that Geo
2
achieves state-of-the-art performance in both localization and synthesis, highlighting the effectiveness of 3D geometric priors for cross-view geo-spatial learning. Our source code can be accessed through https://fobow.github.io/geo2.github.io/.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
2
, a unified framework that leverages Geometric priors from GFMs (e.g., VGGT) to jointly perform Geo-spatial tasks, CVGL and bidirectional CVIS. Despite the 3D reconstruction ability of GFMs, directly applying them to CVGL and CVIS remains challenging due to the large viewpoint gap between ground and aerial imagery. We propose GeoMap, which embeds ground and aerial features into a shared 3D-aware latent space, effectively reducing cross-view discrepancies for localization. This shared latent space naturally bridges cross-view image synthesis in both directions. To exploit this, we propose GeoFlow, a flow-matching model conditioned on geometry-aware latent embeddings. We further introduce a consistency loss to enforce latent alignment between the two synthesis directions, ensuring bidirectional coherence. Extensive experiments on standard benchmarks, including CVUSA, CVACT, and VIGOR, demonstrate that Geo
2
achieves state-of-the-art performance in both localization and synthesis, highlighting the effectiveness of 3D geometric priors for cross-view geo-spatial learning. Our source code can be accessed through https://fobow.github.io/geo2.github.io/.
Kwon, Patrick; Chen, Chen
DreamingComics: A Story Visualization Pipeline via Subject and Layout Customized Generation using Video Models Conference
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026), 2026.
@conference{nokey,
title = {DreamingComics: A Story Visualization Pipeline via Subject and Layout Customized Generation using Video Models},
author = {Patrick Kwon and Chen Chen},
url = {https://arxiv.org/abs/2512.01686},
year = {2026},
date = {2026-03-26},
urldate = {2026-03-26},
publisher = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026)},
abstract = {Current story visualization methods tend to position subjects solely by text and face challenges in maintaining artistic consistency. To address these limitations, we introduce DreamingComics, a layout-aware story visualization framework. We build upon a pretrained video diffusion-transformer (DiT) model, leveraging its spatiotemporal priors to enhance identity and style consistency. For layout-based position control, we propose RegionalRoPE, a region-aware positional encoding scheme that re-indexes embeddings based on the target layout. Additionally, we introduce a masked condition loss to further constrain each subject's visual features to their designated region. To infer layouts from natural language scripts, we integrate an LLM-based layout generator trained to produce comic-style layouts, enabling flexible and controllable layout conditioning. We present a comprehensive evaluation of our approach, showing a 29.2% increase in character consistency and a 36.2% increase in style similarity compared to previous methods, while displaying high spatial accuracy. Our project page is available at this https URL},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Saleem, Muhammad Usama; Patel, Mayur Jagdishbhai; Pinyoanuntapong, Ekkasit; Qin, Zhongxing; Yang, Li; Xue, Hongfei; Helmy, Ahmed; Chen, Chen; Wang, Pu
LiveGesture: Streamable Co-Speech Gesture Generation Model Conference
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026), 2026.
@conference{nokey,
title = {LiveGesture: Streamable Co-Speech Gesture Generation Model},
author = {Muhammad Usama Saleem and Mayur Jagdishbhai Patel and Ekkasit Pinyoanuntapong and Zhongxing Qin and Li Yang and Hongfei Xue and Ahmed Helmy and Chen Chen and Pu Wang},
url = {https://arxiv.org/abs/2604.10927},
year = {2026},
date = {2026-03-26},
urldate = {2026-03-26},
publisher = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026)},
abstract = {We propose LiveGesture, the first fully streamable, speech-driven full-body gesture generation framework that operates with zero look-ahead and supports arbitrary sequence length. Unlike existing co-speech gesture methods, which are designed for offline generation and either treat body regions independently or entangle all joints within a single model, LiveGesture is built from the ground up for causal, region-coordinated motion generation. LiveGesture consists of two main modules: the Streamable Vector Quantized Motion Tokenizer (SVQ) and the Hierarchical Autoregressive Transformer (HAR). The SVQ tokenizer converts the motion sequence of each body region into causal, discrete motion tokens, enabling real-time, streamable token decoding. On top of SVQ, HAR employs region-expert autoregressive (xAR) transformers to model expressive, fine-grained motion dynamics for each body region. A causal spatio-temporal fusion module (xAR Fusion) then captures and integrates correlated motion dynamics across regions. Both xAR and xAR Fusion are conditioned on live, continuously arriving audio signals encoded by a streamable causal audio encoder. To enhance robustness under streaming noise and prediction errors, we introduce autoregressive masking training, which leverages uncertainty-guided token masking and random region masking to expose the model to imperfect, partially erroneous histories during training. Experiments on the BEAT2 dataset demonstrate that LiveGesture produces coherent, diverse, and beat-synchronous full-body gestures in real time, matching or surpassing state-of-the-art offline methods under true zero look-ahead conditions.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Lehyeh, Ayesh Abu; Zhang, Xiaohan; Arrabi, Ahmad; Sultani, Waqas; Chen, Chen; Wshah, Safwan
GeoFlow: Real-Time Fine-Grained Cross-View Geolocalization via Iterative Flow Prediction Conference
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026), 2026.
BibTeX | Links:
@conference{nokey,
title = {GeoFlow: Real-Time Fine-Grained Cross-View Geolocalization via Iterative Flow Prediction},
author = {Ayesh Abu Lehyeh and Xiaohan Zhang and Ahmad Arrabi and Waqas Sultani and Chen Chen and Safwan Wshah},
url = {https://arxiv.org/abs/2603.21943},
year = {2026},
date = {2026-03-23},
urldate = {2026-03-23},
publisher = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026)},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Wang, Zengyan; Mitra, Sirshapan; Modi, Rajat; Lim, Hui Xian Grace; Rawat, Yogesh Singh
Sky2Ground: A Benchmark for Site Modeling under Varying Altitude Conference
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026), 2026.
@conference{nokey,
title = {Sky2Ground: A Benchmark for Site Modeling under Varying Altitude},
author = {Zengyan Wang and Sirshapan Mitra and Rajat Modi and Hui Xian Grace Lim and Yogesh Singh Rawat},
url = {https://arxiv.org/abs/2603.13740},
year = {2026},
date = {2026-03-14},
urldate = {2026-03-14},
publisher = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026)},
abstract = {We introduce Sky2Ground, a three-view dataset designed for varying altitude camera localization, correspondence learning, and reconstruction. The dataset combines structured synthetic imagery with real, in-the-wild images, providing both controlled multi-view geometry and realistic scene noise. Each of the 51 sites contains thousands of satellite, aerial, and ground images spanning wide altitude ranges and nearly orthogonal viewing angles, enabling rigorous evaluation across global-to-local contexts. We benchmark state of the art pose estimation models, including MASt3R, DUSt3R, Map Anything, and VGGT, and observe that the use of satellite imagery often degrades performance, highlighting the challenges under large altitude variations. We also examine reconstruction methods, highlighting the challenges introduced by sparse geometric overlap, varying perspectives, and the use of real imagery, which often introduces noise and reduces rendering quality. To address some of these challenges, we propose SkyNet, a model which enhances cross-view consistency when incorporating satellite imagery with a curriculum-based training strategy to progressively incorporate more satellite views. SkyNet significantly strengthens multi-view alignment and outperforms existing methods by 9.6% on RRA@5 and 18.1% on RTA@5 in terms of absolute performance. Sky2Ground and SkyNet together establish a comprehensive testbed and baseline for advancing large-scale, multi-altitude 3D perception and generalizable camera localization. Code and models will be released publicly for future this http URL page: this https URL},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Azad, Shehreen; Vineet, Vibhav; Rawat, Yogesh Singh
StreamReady: Learning *What* to Answer and *When* in Long Streaming Videos Conference
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026), 2026.
@conference{nokey,
title = {StreamReady: Learning *What* to Answer and *When* in Long Streaming Videos},
author = {Shehreen Azad and Vibhav Vineet and Yogesh Singh Rawat},
url = {https://arxiv.org/abs/2603.08620
https://www.youtube.com/watch?v=WstC36siN-0},
year = {2026},
date = {2026-03-09},
urldate = {2026-03-09},
publisher = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026)},
abstract = {Streaming video understanding often involves time-sensitive scenarios where models need to answer exactly when the supporting visual evidence appears: answering before the evidence reflects speculation, answering after it has passed reduces real-time utility. To capture this behavior, we introduce a readiness-aware formulation of streaming video understanding with the Answer Readiness Score (ARS), a timing-aware objective with asymmetric early and late penalties. When combined with correctness, ARS defines an effective accuracy that measures not just whether a model is right, but whether it answers at the appropriate moment. Building on this formulation, we introduce StreamReady, a framework to unify temporal reasoning with on-time answering through a lightweight readiness mechanism that decides if sufficient evidence has been observed before responding. To evaluate this capability, we further introduce ProReady-QA, a benchmark with annotated answer evidence windows and proactive multi-turn questions across local and global contexts. StreamReady achieves superior performance on ProReady-QA, and consistently outperforms prior methods across eight additional streaming and offline long-video benchmarks, demonstrating robust and broadly generalizable video understanding capability.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Updated 5/4/2026