arXiv subjects

Mozhgan Nasr Azadani

Publications and source records attributed to Mozhgan Nasr Azadani.

7 recordsLinked to original sources

arXiv2026

DIRECT: When and Where Should You Allocate Test-Time Compute in Embodied Planners?

Vision-Language Models (VLMs) are increasingly deployed as high-level planners for embodied agents, with an emerging strategy of scaling test-time compute to improve capability. However, we observe that doing so increases latency, token usage, and FLOPs while yielding uneven, often diminishing gains in downstream success, limiting where embodied agents can be deployed. We argue that choosing when and where to spend test-time compute is central to bringing frontier performance to the real world, and we can budget inference at the planning level. We introduce DIRECT, a routing framework that uses multimodal scene context to allocate compute at each step of a closed-loop agent, selecting a planner for every subtask, improving the success--cost Pareto frontier over fixed model selection. Across three dominant scaling axes, namely chain-of-thought depth, model size, and memory history, our experiments on VLABench and RoboMME show that test-time compute is not a uniform lever: different axes yield qualitatively distinct capability gains. We validate these insights on a physical Franka arm in a DROID setup spanning zero-shot manipulation and long-horizon chaining, where our router matches or exceeds a stronger model's success rate at up to 65% lower average latency. Ultimately, our results show that naively scaling test-time compute is wasteful, and DIRECT can provide frontier-level embodied planning at a fraction of the cost. Project page can be found at jadee-dao.github.io/direct/.

cs.RO↗

arXiv2026

SCOPD: Sparse-Context On-Policy Self-Distillation for Efficient Vision-Language Models

Reasoning vision-language models (VLMs) process images and videos as long sequences of visual tokens, making inference expensive. Training-free token pruning reduces this cost, but aggressive compression can sharply degrade performance, often attributed to irreversible loss of task-relevant visual information. We show that this explanation is incomplete. In a fixed-context Pass@K analysis, repeated sampling from the same pruned visual representation recovers many examples missed by greedy decoding, indicating that useful visual evidence can remain accessible but be used unreliably. We call this the representation-utilization gap. Motivated by this observation, we introduce SCOPD, a sparse-context on-policy self-distillation framework in which a student generates reasoning trajectories from pruned visual tokens while a privileged full-context teacher supervises the same on-policy prefixes. SCOPD requires no ground-truth responses, architectural changes, or additional inference-time computation. We further introduce SCOPD+, which uses a small visual-budget intervention to identify visually sensitive response positions and selectively distill them. At 10% visual-token retention, the Vanilla model retains 86.37% of its unpruned performance across 13 benchmarks. SCOPD raises this to 90.49%, while SCOPD+ further improves it to 92.43%. Across token budgets, benchmarks, and pruning operators, our results show that efficient reasoning depends not only on which visual information survives pruning, but also on how reliably the model learns to use it.

cs.CV↗

arXiv2026

Where Does the Answer Come From? Benchmarking View-Level Visual Evidence Identification in Multi-View MLLMs for Autonomous Driving

Multimodal large language models (MLLMs) achieve strong results on visual reasoning benchmarks, but answer accuracy alone does not indicate whether a model relied on the correct visual evidence. This gap is particularly important in multi-view driving scenes used for autonomous driving, where a model can produce a plausible answer while grounding it in the wrong camera view. We introduce a multi-view visual question answering benchmark for evaluating evidence-source identification: given six synchronized NuScenes views and a question, the model must identify the supporting camera view and answer the question. The benchmark contains 122 conflict-centric question-answer pairs from 73 scenes, spanning causality, counterfactual reasoning, and intent prediction. View labels are proposed by an automatic conflict-mining pipeline and manually verified by annotators. We evaluate three settings: camera-view selection, oracle QA given the golden view, and joint prediction in which the model selects a view and answers in one pass. Answers are evaluated in both multiple-choice and free-form formats, using exact match for structured predictions and an LLM judge for free-form responses. By explicitly separating visual-source identification from answer correctness, the benchmark exposes grounding failures that answer-only evaluation misses.

cs.CL↗

arXiv2026

VISTAQA: Benchmarking Joint Visual Question Answering and Pixel-Level Evidence

Establishing a clear link between model predictions and the visual evidence that supports them is critical for transparency and reliability in multimodal reasoning, yet current multimodal large language model (MLLM) evaluations do not explicitly enforce this alignment. Existing benchmarks assess either textual answer correctness or pixel-level localization in isolation, leaving the coupling of reasoning and grounding an open challenge. We introduce VISTAQA, a comprehensive benchmark for joint evaluation of free-form answer correctness and pixel-level evidence grounding in visual question answering. VISTAQA comprises 1,157 expert-curated samples spanning six task types and six visual domains, ranging from direct perception to compositional and relational reasoning. VISTAQA requires models to not only answer correctly, but to also provide precise segmentation masks that support their answers. It also includes hallucination-aware examples where no valid visual evidence exists. To support this enhanced evaluation, we introduce GROVE, a unified evaluation metric that enforces joint correctness by combining textual accuracy and grounding quality via a per-sample geometric mean, ensuring neither dimension can compensate for deficiencies in the other. Comprehensive experiments across grounding-aware models and hybrid pipelines with general-purpose MLLMs reveal that even the strongest systems achieve limited performance under GROVE, highlighting a substantial gap between answer accuracy and visual evidence alignment.

cs.CV↗

arXiv2026

Rethinking the Mixture of Vision Encoders Paradigm for Enhanced Visual Understanding in Multimodal LLMs

Mixture of Vision Encoders (MoVE) has emerged as a powerful approach to enhance the fine-grained visual understanding of multimodal large language models (MLLMs), improving their ability to handle tasks such as complex optical character recognition and scene understanding. Despite these advances, effectively combining diverse encoders and their visual tokens, while also scaling to high-resolution inputs, remains an open challenge. In this work, we conduct a systematic study of fusion designs for MoVE-based MLLMs, highlighting principles for token-level integration across complementary encoders. Our study shows that a lightweight recipe consisting of post-adaptation fusion with independent projectors, tile-level sequence interleaving, and dynamic tiling with global context delivers strong performance on diverse benchmarks. We integrate these principles into a simple and effective architecture that we call LEO. Extensive evaluation on 11 vision-language benchmarks demonstrates that LEO achieves better results on the majority of tasks compared to existing MoVE-based approaches. Furthermore, LEO adapts effectively to the specialized domain of autonomous driving without altering its architecture or training recipe, achieving competitive performance against established baselines and thereby highlighting its ability to generalize. The code is available at https://github.com/Mozhgan91/LEO.

cs.CV↗

arXiv2025

HAWAII: Hierarchical Visual Knowledge Transfer for Efficient Vision-Language Models

Improving the visual understanding ability of vision-language models (VLMs) is crucial for enhancing their performance across various tasks. While using multiple pretrained visual experts has shown great promise, it often incurs significant computational costs during training and inference. To address this challenge, we propose HAWAII, a novel framework that distills knowledge from multiple visual experts into a single vision encoder, enabling it to inherit the complementary strengths of several experts with minimal computational overhead. To mitigate conflicts among different teachers and switch between different teacher-specific knowledge, instead of using a fixed set of adapters for multiple teachers, we propose to use teacher-specific Low-Rank Adaptation (LoRA) adapters with a corresponding router. Each adapter is aligned with a specific teacher, avoiding noisy guidance during distillation. To enable efficient knowledge distillation, we propose fine-grained and coarse-grained distillation. At the fine-grained level, token importance scores are employed to emphasize the most informative tokens from each teacher adaptively. At the coarse-grained level, we summarize the knowledge from multiple teachers and transfer it to the student using a set of general-knowledge LoRA adapters with a router. Extensive experiments on various vision-language tasks demonstrate the superiority of HAWAII compared to popular open-source VLMs. The code is available at https://github.com/yimuwangcs/wise-hawaii.

cs.CV↗

arXiv2025

LEO-MINI: An Efficient Multimodal Large Language Model using Conditional Token Reduction and Mixture of Multi-Modal Experts

Redundancy of visual tokens in multi-modal large language models (MLLMs) significantly reduces their computational efficiency. Recent approaches, such as resamplers and summarizers, have sought to reduce the number of visual tokens, but at the cost of visual reasoning ability. To address this, we propose LEO-MINI, a novel MLLM that significantly reduces the number of visual tokens and simultaneously boosts visual reasoning capabilities. For efficiency, LEO-MINI incorporates CoTR, a novel token reduction module to consolidate a large number of visual tokens into a smaller set of tokens, using the similarity between visual tokens, text tokens, and a compact learnable query. For effectiveness, to scale up the model's ability with minimal computational overhead, LEO-MINI employs MMoE, a novel mixture of multi-modal experts module. MMOE employs a set of LoRA experts with a novel router to switch between them based on the input text and visual tokens instead of only using the input hidden state. MMoE also includes a general LoRA expert that is always activated to learn general knowledge for LLM reasoning. For extracting richer visual features, MMOE employs a set of vision experts trained on diverse domain-specific data. To demonstrate LEO-MINI's improved efficiency and performance, we evaluate it against existing efficient MLLMs on various benchmark vision-language tasks.

cs.CV↗