arXiv subjects

Pooyan Fazli

Publications and source records attributed to Pooyan Fazli.

At least 19 recordsLinked to original sources

arXiv2026

Video-FLAIR: Not Whether to Reason, But How

Multimodal queries can require different types of reasoning. Some can be answered via perceptual reasoning, extracting information directly from the visual signal, while others require compositional reasoning that combines observations or deliberative reasoning that evaluates competing hypotheses. However, many existing methods apply a uniform reasoning strategy across queries, leading to unnecessary computation on simple tasks and insufficient reasoning on complex ones. We introduce Video-FLAIR, a training framework that learns to select the appropriate reasoning mode for each query using reinforcement learning. During training, the model generates responses under all three modes for the same prompt, enabling direct comparison. A composite reward compares these responses to favor the most effective one based on correctness, grounding, and cost, while discouraging unsupported or misaligned deliberation. This yields a supervision signal for learning adaptive reasoning without per-query annotations. Video-FLAIR improves accuracy over the Qwen2.5-VL base model by +5.4 on MathVista, +4.8 on Video-Holmes, and +4.8 on Video-MMMU, while reducing average token usage to 95 compared to 417 for always-thinking baselines.

cs.CV↗

arXiv2026

VisionPulse: A Virtual Reality System Enabling Accessible Discovery and Navigation for Blind and Low Vision Users

Free exploration is an important aspect of many engaging virtual reality (VR) experiences, yet remains largely inaccessible to blind and low vision (BLV) users due to its reliance on visual feedback. Existing approaches support BLV navigation through prebuilt menus of environment and audio beacons, but offer limited support for free-form discovery. We present VisionPulse, an accessible VR system that enables BLV users to explore virtual environments through natural head and hand movements, combined with auditory, haptic, and text-to-speech feedback. VisionPulse introduces a discovery-driven approach that allows users to progressively uncover regions and objects, alongside navigation support through waypoint guidance and object localization via responsive audio and orientation-based haptics. A study with 12 BLV participants showed a strong preference for VisionPulse's discovery-based exploration and multimodal feedback, without negatively impacting task performance or perceived workload. Our findings underscore the importance of accessible, free-form VR experiences, and contribute insights for inclusive VR design.

cs.HC↗

arXiv2026

ReGATE: Learning Faster and Better with Fewer Tokens in MLLMs

The computational cost of training multimodal large language models (MLLMs) grows rapidly with the number of processed tokens. Existing efficiency methods mainly target inference via token reduction or merging, offering limited benefits during training. We introduce ReGATE (Reference-Guided Adaptive Token Elision), an adaptive token pruning method for accelerating MLLM training. ReGATE adopts a teacher-student framework, in which a frozen teacher LLM provides per-token guidance losses that are fused with an exponential moving average of the student's difficulty estimates. This adaptive scoring mechanism dynamically selects informative tokens while skipping redundant ones in the forward pass, substantially reducing computation without altering the model architecture. Across three representative MLLMs, ReGATE matches the peak accuracy of standard training on MVBench up to 2$\times$ faster, using only 38% of the tokens. With extended training, it even surpasses the baseline across multiple multimodal benchmarks, cutting total token usage by over 41%.

cs.CV↗

arXiv2026

AVATAR: Reinforcement Learning to See, Hear, and Reason Over Video

Multimodal reasoning over long-horizon video is challenging due to the need for precise spatiotemporal fusion and alignment across modalities. While recent methods such as Group Relative Policy Optimization (GRPO) have shown promise in this domain, they suffer from three key limitations: (1) data inefficiency from their on-policy design, (2) a vanishing advantage problem, where identical or near-identical rewards within a group eliminate the learning signal by producing zero-valued advantages, and (3) uniform credit assignment that fails to emphasize critical reasoning steps. We introduce $\textbf{AVATAR}$ ($\textbf{A}$udio-$\textbf{V}$ideo $\textbf{A}$gen$\textbf{t}$ for $\textbf{A}$lignment and $\textbf{R}$easoning), a framework that addresses these limitations through two core components: (1) an off-policy training architecture that improves sample efficiency and resolves vanishing advantages by reusing past experiences with greater reward diversity, and (2) Temporal Advantage Shaping (TAS), a credit assignment strategy that emphasizes early (planning) and late (synthesis) reasoning phases. $\textbf{AVATAR}$ achieves strong performance across various benchmarks, outperforming the Qwen2.5-Omni baseline by $\mathbf{+5.4}$ on MMVU, $\mathbf{+4.9}$ on OmniBench, and $\mathbf{+4.5}$ on Video-Holmes. Furthermore, it surpasses standard GRPO by $\mathbf{+3.7}$ on OmniBench and $\mathbf{+1.9}$ on Video-Holmes, while demonstrating $\textbf{$5$$\times$ sample efficiency}$, requiring $80\%$ fewer generated completions to reach target performance.

cs.CV↗

arXiv2026

ViDscribe: Multimodal AI for Customizing Audio Description and Question Answering in Online Videos

Advances in multimodal large language models enable automatic video narration and question answering (VQA), offering scalable alternatives to labor-intensive, human-authored audio descriptions (ADs) for blind and low vision (BLV) viewers. However, prior AI-driven AD systems rarely adapt to the diverse needs and preferences of BLV individuals across videos and are typically evaluated in controlled, single-session settings. We present ViDscribe, a web-based platform that integrates AI-generated ADs with six types of user customizations and a conversational VQA interface for YouTube videos. Through a longitudinal, in-the-wild study with eight BLV participants, we examine how users engage with customization and VQA features over time. Our results show sustained engagement with both features and that customized ADs improve effectiveness, enjoyment, and immersion compared to default ADs, highlighting the value of personalized, interactive video access for BLV users.

cs.HC↗

arXiv2026

CASHEW: Stabilizing Multimodal Reasoning via Iterative Trajectory Aggregation

Vision-language models achieve strong performance across a wide range of multimodal understanding and reasoning tasks, yet their multi-step reasoning remains unstable. Repeated sampling over the same input often produces divergent reasoning trajectories and inconsistent final predictions. To address this, we introduce two complementary approaches inspired by test-time scaling: (1) CASHEW, an inference-time framework that stabilizes reasoning by iteratively aggregating multiple candidate trajectories into higher-quality reasoning traces, with explicit visual verification filtering hallucinated steps and grounding reasoning in visual evidence, and (2) CASHEW-RL, a learned variant that internalizes this aggregation behavior within a single model. CASHEW-RL is trained using Group Sequence Policy Optimization (GSPO) with a composite reward that encourages correct answers grounded in minimal yet sufficient visual evidence, while adaptively allocating reasoning effort based on task difficulty. This training objective enables robust self-aggregation at inference. Extensive experiments on 13 image understanding, video understanding, and video reasoning benchmarks show significant performance improvements, including gains of up to +26.2 percentage points on ScienceQA and +9.1 percentage points on EgoSchema.

cs.CV↗

arXiv2025

ChartQA-X: Generating Explanations for Visual Chart Reasoning

The ability to explain complex information from chart images is vital for effective data-driven decision-making. In this work, we address the challenge of generating detailed explanations alongside answering questions about charts. We present ChartQA-X, a comprehensive dataset comprising 30,799 chart samples across four chart types, each paired with contextually relevant questions, answers, and explanations. Explanations are generated and selected based on metrics such as faithfulness, informativeness, coherence, and perplexity. Our human evaluation with 245 participants shows that model-generated explanations in ChartQA-X surpass human-written explanations in accuracy and logic and are comparable in terms of clarity and overall quality. Moreover, models fine-tuned on ChartQA-X show substantial improvements across various metrics, including absolute gains of up to 24.57 points in explanation quality, 18.96 percentage points in question-answering accuracy, and 14.75 percentage points on unseen benchmarks for the same task. By integrating explanatory narratives with answers, our approach enables agents to convey complex visual information more effectively, improving comprehension and greater trust in the generated responses.

cs.CV↗

arXiv2025

VidComposition: Can MLLMs Analyze Compositions in Compiled Videos?

The advancement of Multimodal Large Language Models (MLLMs) has enabled significant progress in multimodal understanding, expanding their capacity to analyze video content. However, existing evaluation benchmarks for MLLMs primarily focus on abstract video comprehension, lacking a detailed assessment of their ability to understand video compositions, the nuanced interpretation of how visual elements combine and interact within highly compiled video contexts. We introduce VidComposition, a new benchmark specifically designed to evaluate the video composition understanding capabilities of MLLMs using carefully curated compiled videos and cinematic-level annotations. VidComposition includes 982 videos with 1706 multiple-choice questions, covering various compositional aspects such as camera movement, angle, shot size, narrative structure, character actions and emotions, etc. Our comprehensive evaluation of 33 open-source and proprietary MLLMs reveals a significant performance gap between human and model capabilities. This highlights the limitations of current MLLMs in understanding complex, compiled video compositions and offers insights into areas for further improvement. The leaderboard and evaluation code are available at https://yunlong10.github.io/VidComposition/

cs.CV↗

arXiv2025

EgoVITA: Learning to Plan and Verify for Egocentric Video Reasoning

Egocentric video understanding requires procedural reasoning under partial observability and continuously shifting viewpoints. Current multimodal large language models (MLLMs) struggle with this setting, often generating plausible but visually inconsistent or weakly grounded responses. We introduce $\textbf{EgoVITA}$, a framework that decomposes egocentric video reasoning into a structured $\textit{plan-then-verify}$ process. The model first generates an $\textbf{egocentric plan}$: a causal sequence of anticipated actions from a first-person perspective. This plan is then evaluated by an $\textbf{exocentric verification}$ stage that uses third-person reasoning over the same video to verify its spatiotemporal and logical consistency, without exocentric video input. This decomposition enables cross-perspective feedback without requiring paired ego-exo supervision. To train this reasoning process, we adopt Group Relative Policy Optimization (GRPO) with two dense reward signals: one that grounds anticipated actions in subsequent visual observations and another that reinforces consistent third-person verification. $\textbf{EgoVITA}$ achieves state-of-the-art performance on egocentric reasoning benchmarks, outperforming Qwen2.5-VL-7B by $\mathbf{+7.7}$ on EgoBlind and $\mathbf{+4.4}$ on EgoOrient, while maintaining strong generalization on exocentric video tasks with only $52k$ training samples.

cs.CV↗

arXiv2025

FrameOracle: Learning What to See and How Much to See in Videos

Vision-language models (VLMs) advance video understanding but operate under tight computational budgets, making performance dependent on selecting a small, high-quality subset of frames. Existing frame sampling strategies, such as uniform or fixed-budget selection, fail to adapt to variations in content density or task complexity. To address this, we present FrameOracle, a lightweight, plug-and-play module that predicts both (1) which frames are most relevant to a given query and (2) how many frames are needed. FrameOracle is trained via a curriculum that progresses from weak proxy signals, such as cross-modal similarity, to stronger supervision with FrameOracle-41K, the first large-scale VideoQA dataset with validated keyframe annotations specifying minimal sufficient frames per question. Extensive experiments across five VLMs and six benchmarks show that FrameOracle reduces 16-frame inputs to an average of 10.4 frames without accuracy loss. When starting from 64-frame candidates, it reduces inputs to 13.9 frames on average while improving accuracy by 1.5%, achieving state-of-the-art efficiency-accuracy trade-offs for scalable video understanding.

cs.CV↗

arXiv2025

VideoPASTA: 7K Preference Pairs That Matter for Video-LLM Alignment

Video-language models (Video-LLMs) excel at understanding video content but struggle with spatial relationships, temporal ordering, and cross-frame continuity. To address these limitations, we introduce VideoPASTA (Preference Alignment with Spatio-Temporal-Cross Frame Adversaries), a framework that enhances Video-LLMs through targeted preference optimization. VideoPASTA trains models to distinguish accurate video representations from carefully crafted adversarial examples that deliberately violate spatial, temporal, or cross-frame relationships. With only 7,020 preference pairs and Direct Preference Optimization, VideoPASTA enables models to learn robust representations that capture fine-grained spatial details and long-range temporal dynamics. Experiments demonstrate that VideoPASTA is model agnostic and significantly improves performance, for example, achieving gains of up to +3.8 percentage points on LongVideoBench, +4.1 on VideoMME, and +4.0 on MVBench, when applied to various state-of-the-art Video-LLMs. These results demonstrate that targeted alignment, rather than massive pretraining or architectural modifications, effectively addresses core video-language challenges. Notably, VideoPASTA achieves these improvements without any human annotation or captioning, relying solely on 32-frame sampling. This efficiency makes our approach a scalable plug-and-play solution that seamlessly integrates with existing models while preserving their original capabilities.

cs.CV↗

arXiv2025

VideoSAVi: Self-Aligned Video Language Models without Human Supervision

Recent advances in video-large language models (Video-LLMs) have led to significant progress in video understanding. Current preference optimization methods often rely on proprietary APIs or human-annotated captions to generate preference data (i.e., pairs of model outputs ranked by quality or alignment with human judgment), which is then used to train models for video-language alignment. This approach is both costly and labor-intensive. To address this limitation, we introduce VideoSAVi (Self-Aligned Video Language Model), a self-training pipeline that enables Video-LLMs to learn from video content without external supervision. Our approach includes a self-critiquing mechanism that identifies reasoning errors in the model's initial responses and generates improved alternatives, creating preference pairs directly from video content. VideoSAVi then applies Direct Preference Optimization (DPO) to iteratively train the model using the preference data, thus enhancing its temporal and spatial reasoning for video understanding. Experiments show that VideoSAVi delivers significant improvements across multiple benchmarks, including a +4.2 percentage point gain on MVBench, +3.9 on PerceptionTest, and +6.8 on the challenging EgoSchema dataset compared to baseline models. Our model-agnostic approach is computationally efficient, requiring only 32 frames, offering a promising direction for self-aligned video understanding without reliance on external models or annotations.

cs.CV↗

arXiv2025

DescribePro: Collaborative Audio Description with Human-AI Interaction

Audio description (AD) makes video content accessible to millions of blind and low vision (BLV) users. However, creating high-quality AD involves a trade-off between the precision of human-crafted descriptions and the efficiency of AI-generated ones. To address this, we present DescribePro a collaborative AD authoring system that enables describers to iteratively refine AI-generated descriptions through multimodal large language model prompting and manual editing. DescribePro also supports community collaboration by allowing users to fork and edit existing ADs, enabling the exploration of different narrative styles. We evaluate DescribePro with 18 describers (9 professionals and 9 novices) using quantitative and qualitative methods. Results show that AI support reduces repetitive work while helping professionals preserve their stylistic choices and easing the cognitive load for novices. Collaborative tags and variations show potential for providing customizations, version control, and training new describers. These findings highlight the potential of collaborative, AI-assisted tools to enhance and scale AD authorship.

cs.HC↗

arXiv2025

Describe Now: User-Driven Audio Description for Blind and Low Vision Individuals

Audio descriptions (AD) make videos accessible for blind and low vision (BLV) users by describing visual elements that cannot be understood from the main audio track. AD created by professionals or novice describers is time-consuming and offers little customization or control to BLV viewers on description length and content and when they receive it. To address this gap, we explore user-driven AI-generated descriptions, enabling BLV viewers to control both the timing and level of detail of the descriptions they receive. In a study, 20 BLV participants activated audio descriptions for seven different video genres with two levels of detail: concise and detailed. Our findings reveal differences in the preferred frequency and level of detail of ADs for different videos, participants' sense of control with this style of AD delivery, and its limitations. We discuss the implications of these findings for the development of future AD tools for BLV users.

cs.HC↗

arXiv2025

VidHalluc: Evaluating Temporal Hallucinations in Multimodal Large Language Models for Video Understanding

Multimodal large language models (MLLMs) have recently shown significant advancements in video understanding, excelling in content reasoning and instruction-following tasks. However, hallucination, where models generate inaccurate or misleading content, remains underexplored in the video domain. Building on the observation that MLLM visual encoders often fail to distinguish visually different yet semantically similar video pairs, we introduce VidHalluc, the largest benchmark designed to examine hallucinations in MLLMs for video understanding. It consists of 5,002 videos, paired to highlight cases prone to hallucinations. VidHalluc assesses hallucinations across three critical dimensions: (1) action, (2) temporal sequence, and (3) scene transition. Comprehensive testing shows that most MLLMs are vulnerable to hallucinations across these dimensions. Furthermore, we propose DINO-HEAL, a training-free method that reduces hallucinations by incorporating spatial saliency from DINOv2 to reweight visual features during inference. Our results show that DINO-HEAL consistently improves performance on VidHalluc, achieving an average improvement of 3.02% in mitigating hallucinations across all tasks. Both the VidHalluc benchmark and DINO-HEAL code are available at https://people-robots.github.io/vidhalluc.

cs.CV↗

arXiv2025

VideoA11y: Method and Dataset for Accessible Video Description

Video descriptions are crucial for blind and low vision (BLV) users to access visual content. However, current artificial intelligence models for generating descriptions often fall short due to limitations in the quality of human annotations within training datasets, resulting in descriptions that do not fully meet BLV users' needs. To address this gap, we introduce VideoA11y, an approach that leverages multimodal large language models (MLLMs) and video accessibility guidelines to generate descriptions tailored for BLV individuals. Using this method, we have curated VideoA11y-40K, the largest and most comprehensive dataset of 40,000 videos described for BLV users. Rigorous experiments across 15 video categories, involving 347 sighted participants, 40 BLV participants, and seven professional describers, showed that VideoA11y descriptions outperform novice human annotations and are comparable to trained human annotations in clarity, accuracy, objectivity, descriptiveness, and user satisfaction. We evaluated models on VideoA11y-40K using both standard and custom metrics, demonstrating that MLLMs fine-tuned on this dataset produce high-quality accessible descriptions. Code and dataset are available at https://people-robots.github.io/VideoA11y.

cs.CV↗

arXiv2024

OSCaR: Object State Captioning and State Change Representation

The capability of intelligent models to extrapolate and comprehend changes in object states is a crucial yet demanding aspect of AI research, particularly through the lens of human interaction in real-world settings. This task involves describing complex visual environments, identifying active objects, and interpreting their changes as conveyed through language. Traditional methods, which isolate object captioning and state change detection, offer a limited view of dynamic environments. Moreover, relying on a small set of symbolic words to represent changes has restricted the expressiveness of the language. To address these challenges, in this paper, we introduce the Object State Captioning and State Change Representation (OSCaR) dataset and benchmark. OSCaR consists of 14,084 annotated video segments with nearly 1,000 unique objects from various egocentric video collections. It sets a new testbed for evaluating multimodal large language models (MLLMs). Our experiments demonstrate that while MLLMs show some skill, they lack a full understanding of object state changes. The benchmark includes a fine-tuned model that, despite initial capabilities, requires significant improvements in accuracy and generalization ability for effective understanding of these changes. Our code and dataset are available at https://github.com/nguyennm1024/OSCaR.

cs.CV↗

arXiv2023

CLearViD: Curriculum Learning for Video Description

Video description entails automatically generating coherent natural language sentences that narrate the content of a given video. We introduce CLearViD, a transformer-based model for video description generation that leverages curriculum learning to accomplish this task. In particular, we investigate two curriculum strategies: (1) progressively exposing the model to more challenging samples by gradually applying a Gaussian noise to the video data, and (2) gradually reducing the capacity of the network through dropout during the training process. These methods enable the model to learn more robust and generalizable features. Moreover, CLearViD leverages the Mish activation function, which provides non-linearity and non-monotonicity and helps alleviate the issue of vanishing gradients. Our extensive experiments and ablation studies demonstrate the effectiveness of the proposed model. The results on two datasets, namely ActivityNet Captions and YouCook2, show that CLearViD significantly outperforms existing state-of-the-art models in terms of both accuracy and diversity metrics.

cs.CV↗