A blank-image ablation test reveals that high probe accuracy on VLM spatial reasoning frequently reflects priors or inverted signs rather than image grounding, with horizontal grounded, vertical prior, and depth inverted.
mega hub Mixed citations
Qwen2.5-VL Technical Report
Mixed citation behavior. Most common role is background (53%).
1318
Pith papers citing it
53
external citations · Pith
Background
53% of classified citations
abstract
We introduce Qwen2.5-VL, the latest flagship model of Qwen vision-language series, which demonstrates significant advancements in both foundational capabilities and innovative functionalities. Qwen2.5-VL achieves a major leap forward in understanding and interacting with the world through enhanced visual recognition, precise object localization, robust document parsing, and long-video comprehension. A standout feature of Qwen2.5-VL is its ability to localize objects using bounding boxes or points accurately. It provides robust structured data extraction from invoices, forms, and tables, as well as detailed analysis of charts, diagrams, and layouts. To handle complex inputs, Qwen2.5-VL introduces dynamic resolution processing and absolute time encoding, enabling it to process images of varying sizes and videos of extended durations (up to hours) with second-level event localization. This allows the model to natively perceive spatial scales and temporal dynamics without relying on traditional normalization techniques. By training a native dynamic-resolution Vision Transformer (ViT) from scratch and incorporating Window Attention, we reduce computational overhead while maintaining native resolution. As a result, Qwen2.5-VL excels not only in static image and document understanding but also as an interactive visual agent capable of reasoning, tool usage, and task execution in real-world scenarios such as operating computers and mobile devices. Qwen2.5-VL is available in three sizes, addressing diverse use cases from edge AI to high-performance computing. The flagship Qwen2.5-VL-72B model matches state-of-the-art models like GPT-4o and Claude 3.5 Sonnet, particularly excelling in document and diagram understanding. Additionally, Qwen2.5-VL maintains robust linguistic performance, preserving the core language competencies of the Qwen2.5 LLM.
hub tools
citation-role summary
background 153
method 58
baseline 57
dataset 5
other 3
citation-polarity summary
claims ledger
- abstract We introduce Qwen2.5-VL, the latest flagship model of Qwen vision-language series, which demonstrates significant advancements in both foundational capabilities and innovative functionalities. Qwen2.5-VL achieves a major leap forward in understanding and interacting with the world through enhanced visual recognition, precise object localization, robust document parsing, and long-video comprehension. A standout feature of Qwen2.5-VL is its ability to localize objects using bounding boxes or points accurately. It provides robust structured data extraction from invoices, forms, and tables, as wel
authors
mega hub controls
export citing contexts JSON
export graph JSON
export full bundle JSON
open full Pith review
annotated reader queued
Recognition alignment
counterfactual ablation
co-cited works
representative citing papers
DataComp-VLM benchmark shows instruction-heavy data mixing outperforms filtering for VLM training, with DCVLM-Baseline achieving 63.6% on 33 tasks for 8B models (+5.4pp over FineVision).
Introduces the first large-scale multimodal benchmark MedLayXPlain-122K showing medical VLMs suffer significant lay-register degradation while general VLMs lack clinical precision.
A safety direction estimated in a source LLM is transported to a target generator through lightweight alignment on benign data alone, matching native safety performance without any target-side unsafe data.
The paper presents RoboAbstention, a new benchmark showing frontier VLMs and embodied planners abstain on only 16.5-39% of 6,069 instructions grounded in robotics images, with prompting interventions raising rates to 88-93% but not solving the problem.
A multi-view transformer predicts dense perspective fields that feed a geometric optimizer to estimate camera intrinsics and gravity from arbitrary numbers of real-world views.
An MLLM-guided architecture with a mixture of frequency experts and relational alignment loss achieves state-of-the-art all-in-one image restoration, outperforming prior methods by up to 1.35 dB on the CDD11 dataset.
Hilbert-Geo creates the first unified formal language for solid geometry and a two-step parsing-then-reasoning method that reaches SOTA accuracy on solid geometry benchmarks.
M³Att poisons medical multimodal RAG by pairing covert textual misinformation with query-agnostic visual perturbations that increase retrieval of the bad content, causing LLMs to generate clinically plausible but incorrect responses.
DeepTumorVQA is a new stage-wise 3D CT VQA benchmark showing that quantitative measurement is the main failure point for current medical VLMs and that tool augmentation substantially improves later reasoning stages.
Flame3D enables zero-shot compositional 3D scene reasoning by representing scenes as editable visual-textual memories exposed to agentic MLLMs through composable and synthesizable spatial tools.
RuleSafe-VL creates 2,166 rule-conditioned cases from 93 atomic rules and 92 relations across three policy families to diagnose where VLMs fail at rule-based content moderation reasoning.
MedHorizon benchmark reveals current multimodal LLMs achieve only 41.1% accuracy on long medical videos due to failures in sparse evidence retrieval and procedural reasoning.
EgoSound is a new benchmark with 7315 QA pairs across seven tasks to evaluate egocentric sound understanding in multimodal large language models.
VLRS-Bench is the first benchmark dedicated to complex vision-language reasoning in remote sensing, with 2000 QA pairs across 14 tasks in cognition, decision, and prediction dimensions.
Cornfigurator is the first automated deployment planner for generic any-to-any multimodal models that explores the full range of colocation-to-disaggregation strategies and delivers 1.12x to 6.32x higher goodput than existing systems or expert plans.
ToG-Bench is the first benchmark for task-oriented spatio-temporal video grounding in egocentric videos, with explicit-implicit dual grounding and one-to-many object scenarios across 100 ScanNet clips and 2704 instructions.
FLEX is the first large-scale multimodal multiview dataset for fitness AQA, featuring RGB, 3D pose, sEMG and physiological data plus a Fitness Knowledge Graph for structured annotations and a VideoQA benchmark.
LePaX enables high-resolution chest X-ray report generation by learning to allocate resolution to diagnostically relevant regions and fusing high-res patches back into global features without increasing token count.
EgoPolice introduces a 185-hour annotated police body-worn camera benchmark showing state-of-the-art video models fail on high-stakes actions due to motion, occlusion, and low inter-class visual separability.
TILDE derives a minimum-deviation, energy-tilted target distribution for concept unlearning in diffusion models and realizes it via residual ∇-GFlowNet training.
MTLA is a training-free, post-hoc confidence score for multimodal LLM localization that restricts attention aggregation to the model's own predicted region and tokens, substantially improving hallucination detection and re-ranking across image, video, and audio.
ReQuest introduces an uncertainty-driven question-adaptive keyframe selector with rethinking routing and adaptive NMS that boosts long-form video QA accuracy on Video-MME, MLVU, and LongVideoBench without fine-tuning the base MLLM.
P2R decouples perception from reasoning in VLMs via a two-stage process and PRA-GRPO alternating RL training, reporting gains such as 93.2% on V-Star for the 4B model over its Qwen3-VL backbone.
citing papers explorer
-
Decodable Is Not Grounded: A Vision-Ablation Arbiter for VLM Spatial Reasoning
A blank-image ablation test reveals that high probe accuracy on VLM spatial reasoning frequently reflects priors or inverted signs rather than image grounding, with horizontal grounded, vertical prior, and depth inverted.
-
DataComp-VLM: Improved Open Datasets for Vision-Language Models
DataComp-VLM benchmark shows instruction-heavy data mixing outperforms filtering for VLM training, with DCVLM-Baseline achieving 63.6% on 33 tasks for 8B models (+5.4pp over FineVision).
-
MEDLAYXPLAIN: Benchmarking the Expert-Lay Gap in Medical Vision-Language Models
Introduces the first large-scale multimodal benchmark MedLayXPlain-122K showing medical VLMs suffer significant lay-register degradation while general VLMs lack clinical precision.
-
Do Models Share Safety Representations? Cross-Model Steering for Safe Visual Generation
A safety direction estimated in a source LLM is transported to a target generator through lightweight alignment on benign data alone, matching native safety performance without any target-side unsafe data.
-
The Yes-Man Syndrome: Benchmarking Abstention in Embodied Robotic Agents
The paper presents RoboAbstention, a new benchmark showing frontier VLMs and embodied planners abstain on only 16.5-39% of 6,069 instructions grounded in robotics images, with prompting interventions raising rates to 88-93% but not solving the problem.
-
CalibAnyView: Beyond Single-View Camera Calibration in the Wild
A multi-view transformer predicts dense perspective fields that feed a geometric optimizer to estimate camera intrinsics and gravity from arbitrary numbers of real-world views.
-
Leveraging Multimodal Large Language Models for All-in-One Image Restoration via a Mixture of Frequency Experts
An MLLM-guided architecture with a mixture of frequency experts and relational alignment loss achieves state-of-the-art all-in-one image restoration, outperforming prior methods by up to 1.35 dB on the CDD11 dataset.
-
Hilbert-Geo: Solving Solid Geometric Problems by Neural-Symbolic Reasoning
Hilbert-Geo creates the first unified formal language for solid geometry and a two-step parsing-then-reasoning method that reaches SOTA accuracy on solid geometry benchmarks.
-
Knowledge Poisoning Attacks on Medical Multi-Modal Retrieval-Augmented Generation
M³Att poisons medical multimodal RAG by pairing covert textual misinformation with query-agnostic visual perturbations that increase retrieval of the bad content, causing LLMs to generate clinically plausible but incorrect responses.
-
DeepTumorVQA: A Hierarchical 3D CT Benchmark for Stage-Wise Evaluation of Medical VLMs and Tool-Augmented Agents
DeepTumorVQA is a new stage-wise 3D CT VQA benchmark showing that quantitative measurement is the main failure point for current medical VLMs and that tool augmentation substantially improves later reasoning stages.
-
Flame3D: Zero-shot Compositional Reasoning of 3D Scenes with Agentic Language Models
Flame3D enables zero-shot compositional 3D scene reasoning by representing scenes as editable visual-textual memories exposed to agentic MLLMs through composable and synthesizable spatial tools.
-
RuleSafe-VL: Evaluating Rule-Conditioned Decision Reasoning in Vision-Language Content Moderation
RuleSafe-VL creates 2,166 rule-conditioned cases from 93 atomic rules and 92 relations across three policy families to diagnose where VLMs fail at rule-based content moderation reasoning.
-
MedHorizon: Towards Long-context Medical Video Understanding in the Wild
MedHorizon benchmark reveals current multimodal LLMs achieve only 41.1% accuracy on long medical videos due to failures in sparse evidence retrieval and procedural reasoning.
-
EgoSound: Benchmarking Sound Understanding in Egocentric Videos
EgoSound is a new benchmark with 7315 QA pairs across seven tasks to evaluate egocentric sound understanding in multimodal large language models.
-
VLRS-Bench: A Vision-Language Reasoning Benchmark for Remote Sensing
VLRS-Bench is the first benchmark dedicated to complex vision-language reasoning in remote sensing, with 2000 QA pairs across 14 tasks in cognition, decision, and prediction dimensions.
-
Cornfigurator: Automated Planning for Any-to-Any Multimodal Model Serving
Cornfigurator is the first automated deployment planner for generic any-to-any multimodal models that explores the full range of colocation-to-disaggregation strategies and delivers 1.12x to 6.32x higher goodput than existing systems or expert plans.
-
ToG-Bench: Task-Oriented Spatio-Temporal Grounding in Egocentric Videos
ToG-Bench is the first benchmark for task-oriented spatio-temporal video grounding in egocentric videos, with explicit-implicit dual grounding and one-to-many object scenarios across 100 ScanNet clips and 2704 instructions.
-
FLEX: A Largescale Multimodal, Multiview Dataset for Learning Structured Representations for Fitness Action Quality Assessment
FLEX is the first large-scale multimodal multiview dataset for fitness AQA, featuring RGB, 3D pose, sEMG and physiological data plus a Fitness Knowledge Graph for structured annotations and a VideoQA benchmark.
-
Seeing What Matters: Lesion-Aware High-Resolution Patch Discovery and Fusion for Chest X-ray Report Generation
LePaX enables high-resolution chest X-ray report generation by learning to allocate resolution to diagnostically relevant regions and fusing high-res patches back into global features without increasing token count.
-
EgoPolice: A Benchmark for Egocentric Video Understanding in High-Stakes Police Body-Worn Camera Footage
EgoPolice introduces a 185-hour annotated police body-worn camera benchmark showing state-of-the-art video models fail on high-stakes actions due to motion, occlusion, and low inter-class visual separability.
-
TILDE: TILt-based Distributional Erasure for Concept Unlearning
TILDE derives a minimum-deviation, energy-tilted target distribution for concept unlearning in diffusion models and realizes it via residual ∇-GFlowNet training.
-
Propose and Attend: Training-free MLLM Grounding Confidence via Multi-Token Localized Attention
MTLA is a training-free, post-hoc confidence score for multimodal LLM localization that restricts attention aggregation to the model's own predicted region and tokens, substantially improving hallucination detection and re-ranking across image, video, and audio.
-
ReQuest: Rethinking-based Question-Aware Frame Selection for Long-Form Video QA
ReQuest introduces an uncertainty-driven question-adaptive keyframe selector with rethinking routing and adaptive NMS that boosts long-form video QA accuracy on Video-MME, MLVU, and LongVideoBench without fine-tuning the base MLLM.
-
Perceive-to-Reason: Decoupling Perception and Reasoning for Fine-Grained Visual Reasoning
P2R decouples perception from reasoning in VLMs via a two-stage process and PRA-GRPO alternating RL training, reporting gains such as 93.2% on V-Star for the 4B model over its Qwen3-VL backbone.
-
MoHallBench: A Benchmark for Motion Hallucination in Video Large Language Models
MoHallBench is a new benchmark evaluating motion hallucination in VideoLLMs from co-occurrence priors, sequential inference, and similarity confusion, revealing decoupling from action recognition performance.
-
LongVQUBench: Benchmarking Long-Term Video Quality Understanding of Vision-Language Models
LongVQUBench introduces a hierarchical benchmark with local, cross-event, and global quality understanding tasks plus needle distortion QA to measure LVLMs' long-term video quality reasoning.
-
TrajLoc: Trajectory-Attention Localization for Multi-Object Motion Control
TrajLoc enforces per-object trajectory constraints in I2V generation via attention-layer Gaussian heatmap substitution, yielding +4.3 dB PSNR and 51% lower endpoint error on datasets with up to 20 objects across two backbones.
-
Learning to Watch: Active Video Anomaly Understanding via Interleaved Policy Optimization
Introduces Anom-π framework for active video anomaly understanding via interleaved policy optimization and iDPO under weak supervision, claiming a 2B model outperforms larger SOTA VAU models.
-
No Place to Hide: Benchmarking Video Hallucination with Background-Controlled Pairs
Introduces VidPair-Halluc benchmark of 1K background-controlled adversarial video pairs and 11K QA pairs generated via PairFlow pipeline to evaluate hallucination in LVMs.
-
Attend, Transform, or Silence: Operator-Level Visual Skipping for Efficient Multimodal LLM Inference
The paper proposes an operator-level visual-token skipping framework for MLLMs that reduces TFLOPs by 33.7% on Qwen3-VL while retaining 99.5% performance across VQA benchmarks.
-
Arena-T2I Hard: Benchmarking and Improving Faithfulness with Dependency-Aware Checklist
Arena-T2I Hard benchmark with ~30 decomposed constraints per prompt and a dependency-aware checklist reward yields better faithfulness-aesthetics trade-off than single-reward or weighted-sum baselines on SD3.5-Medium and FLUX.1-dev.
-
Visual Semantic Entropy: Do Vision Language Models Recognize Visual Ambiguity?
VSE perturbs images only to probe visual ambiguity in VLMs, clusters outputs into semantic prototypes, and computes mass-weighted dispersion, outperforming prior entropy methods on five VQA benchmarks across five models.
-
Learning to Deny: Action Denial in Multimodal Large Language Models
MLLMs drop from over 85% accuracy on action presence to under 50% on matched action-denial videos, exposing a causal verification gap that causal graph prompts partially close.
-
OmniCoT: A Benchmark for Global and Multi-Step Panoramic Reasoning
OmniCoT is a new panoramic reasoning benchmark with 6.7K eval, 1K real, and 14.3K training examples plus a two-stage SFT+GRPO training method to enforce global 360-degree consistency.
-
On Test-Time Scaling for Vision-Language Models
Small well-performing LVLMs gain the largest benefits from test-time scaling (up to ~30% improvement), often matching or exceeding larger models, while visual tokens contribute mainly early in the reasoning chain.
-
OrthoTryOn: Geometric Orthogonalization for Conflict-Free Unified Fashion Generation
OrthoTryOn uses Orthogonal Subspace Projection on shared LoRA and Fisher-guided Negative Guidance to enable conflict-free unified fashion generation, outperforming task-specific models on benchmarks.
-
Video-MME-Logical: A Controlled Diagnostic Benchmark for Video Temporal-Logical Reasoning
Introduces Video-MME-Logical benchmark for controlled diagnostic evaluation of temporal-logical reasoning in MLLMs via five operations and 25 fine-grained tasks.
-
DiCoBench: Benchmarking Multi-Image Fine-Grained Perception via Differential and Commonality Visual Cues
DiCoBench is a new high-resolution multi-image benchmark exposing large gaps between top MLLMs and human performance (98.3%) on differential and commonality visual cue perception.
-
GeMoE: Gating Entropy is All You Need for Uncertainty-aware Adaptive Routing in MoE-based Large Vision-Language Models
GeMoE adaptively sets the number of experts per token via gating entropy, retaining 99.5% of static-routing performance while raising average sparsity by 36.5%.
-
TriViewBench: Controlled Complexity Scaling for Multi-View Structural Reasoning in MLLMs
TriViewBench shows all tested MLLMs follow the same capability order with sharp drops on complex multi-view tasks and near-zero gain from Chain-of-Thought prompting.
-
In-Context World Modeling for Robotic Control
Prepending a few self-generated random interaction clips as context lets VLA policies identify novel camera viewpoints and morphologies at test time and outperform multi-view baselines without parameter updates.
-
SSMNBench: Diagnosing Image-based Cross-View Human-Object Understanding via Single-View Sufficiency and Multi-View Necessity
SSMNBench shows that MLLMs suffer distraction degradation on single-view-sufficient tasks and fail to integrate geometric evidence across views, instead relying on semantic averaging and view preference.
-
FeVOS: Foresight Expression Video Object Segmentation
Introduces the FeVOS task, a 968-clip dataset with foresight expressions, and an MLLM model FeVOS-R1 trained via SFT then RL that reports SOTA on the new task plus generalization to prior RVOS benchmarks.
-
C3-Bench: A Context-Aware Change Captioning Benchmark
C3-Bench supplies a multi-domain dataset and LLM-based evaluation protocol that exposes systematic failures in existing change captioning models outside their training regimes.
-
EgoSAT: A Comprehensive Benchmark of Egocentric Streaming Interaction Understanding
EgoSAT is the first benchmark unifying retrospective, online, and prospective reasoning tasks in egocentric streaming video to evaluate VLMs, revealing struggles with temporal modeling and mis-calibration.
-
Ill-Posed by Design: Probing Evidence Use in VLMs
VLMs rely primarily on target identity for monocular size estimation and largely ignore scene geometry, trailing text-only LLMs on in-the-wild data.
-
ChartWalker: Benchmarking the Cross-Chart RAG Task with Hierarchical Knowledge Graphs
ChartWalker provides a hierarchical knowledge graph construction method and structure-aware sampling to generate cross-chart RAG benchmarks, releasing ChartWalker-Bench that exposes performance gaps across RAG paradigms.
-
Keep The Essentials: Efficient Reference Conditioned Generation via Token Dropping
Sparse Context achieves 2-4x faster inference in reference-conditioned diffusion models by fine-tuning with random token dropping and applying task-aware selection at inference time, without loss of visual quality.
-
Flow as Flow: Modeling Robot Velocity Fields as Probability Velocity Fields for Flow-Based Object Manipulation
Flow as Flow models robot flows as probability flows using flow matching to generate velocity fields more efficiently than prior sparse keypoint approaches.
-
Each Judge Its Own Yardstick: Discovering Per-VLM Taxonomies for Physical Video Evaluation
JudgeFit produces per-VLM physical video evaluation taxonomies that improve held-out accuracy by a mean 32% relative to a single global schema across 16 models from eight families.