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).
super hub Mixed citations
Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution
Mixed citation behavior. Most common role is background (60%).
667
Pith papers citing it
Background
60% of classified citations
abstract
We present the Qwen2-VL Series, an advanced upgrade of the previous Qwen-VL models that redefines the conventional predetermined-resolution approach in visual processing. Qwen2-VL introduces the Naive Dynamic Resolution mechanism, which enables the model to dynamically process images of varying resolutions into different numbers of visual tokens. This approach allows the model to generate more efficient and accurate visual representations, closely aligning with human perceptual processes. The model also integrates Multimodal Rotary Position Embedding (M-RoPE), facilitating the effective fusion of positional information across text, images, and videos. We employ a unified paradigm for processing both images and videos, enhancing the model's visual perception capabilities. To explore the potential of large multimodal models, Qwen2-VL investigates the scaling laws for large vision-language models (LVLMs). By scaling both the model size-with versions at 2B, 8B, and 72B parameters-and the amount of training data, the Qwen2-VL Series achieves highly competitive performance. Notably, the Qwen2-VL-72B model achieves results comparable to leading models such as GPT-4o and Claude3.5-Sonnet across various multimodal benchmarks, outperforming other generalist models. Code is available at https://github.com/QwenLM/Qwen2-VL .
hub tools
citation-role summary
background 103
baseline 28
method 26
dataset 6
other 2
citation-polarity summary
claims ledger
- abstract We present the Qwen2-VL Series, an advanced upgrade of the previous Qwen-VL models that redefines the conventional predetermined-resolution approach in visual processing. Qwen2-VL introduces the Naive Dynamic Resolution mechanism, which enables the model to dynamically process images of varying resolutions into different numbers of visual tokens. This approach allows the model to generate more efficient and accurate visual representations, closely aligning with human perceptual processes. The model also integrates Multimodal Rotary Position Embedding (M-RoPE), facilitating the effective fusion
authors
co-cited works
representative citing papers
NeuroQA is a large-scale 3D brain MRI visual question answering benchmark with verified image-grounded QA pairs, multi-domain coverage, and baseline evaluations showing current models lag behind text-only performance.
MI-CXR is a new benchmark that shows state-of-the-art vision-language models achieve only 29.3% accuracy on longitudinal reasoning tasks across multi-visit chest X-ray sequences.
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.
GLADOS reconstructs 3D geometry from disjoint views by generating intermediate perspectives, performing robust coarse alignment that tolerates generative inconsistencies, and iteratively expanding context for consistency.
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.
CADFS supplies a large real-world CAD dataset and FeatureScript representation that, after VLM fine-tuning, produces more accurate and feature-rich designs than prior generative CAD systems.
SpikeMLLM is the first spike-based MLLM framework that maintains near-lossless performance under aggressive timestep compression and delivers 9x throughput and 25x power efficiency gains via a custom RTL accelerator.
VLM-UnBench demonstrates that prompt-based training-free unlearning in VLMs leaves forget accuracy near the no-instruction baseline except under oracle conditions that reveal the target concept.
VAREX benchmark shows structured output compliance limits models under 4B parameters more than extraction ability, with layout-preserving text giving the largest accuracy gains over images.
Molmo2 delivers state-of-the-art open-weight video VLMs with new grounding datasets and training methods that outperform prior open models and match or exceed some proprietary ones on pointing and tracking tasks.
Introduces a method to design structure-specific relational inductive biases for a base transformer architecture, enabling end-to-end transcription of documents with intrinsic structures, demonstrated on sheet music, shape drawings, and mechanical engineering drawings.
Molmo VLMs trained on newly collected PixMo open datasets achieve state-of-the-art performance among open-weight models and surpass multiple proprietary VLMs including Claude 3.5 Sonnet and Gemini 1.5 Pro.
Introduces APRS task and PanoSeeker agent using VLM plus EgoSphere memory for active 360° search and segmentation, outperforming baselines on a new benchmark.
Proposes WUICC task and WUICC-bench dataset, then evaluates 11 image difference captioning methods plus 2 LLMs on web UI changes.
VLMs show chance-level depth ordering performance (47-56%) on controlled images, driven by language bias rather than pictorial cues, with no improvement from CoT or ICL.
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.
TopoGPT pre-trains an autoregressive transformer on serialized lane graphs from 3.3M scenes to learn geometry priors and uses a perception adapter to apply it to BEV features for improved lane graph prediction on OpenLane-V2.
RCT dataset with sequence-preserving splits demonstrates that tactile-to-text models achieve only 25.1% Recall@1 on held-out materials, exposing generalization as the core challenge.
RRG trains MLLMs via a reinforced multimodal reference game with contrastive rewards on hard positives and negatives to produce accurate, discriminative concept descriptions, achieving SOTA on personalization benchmarks.
VLMs default to visual grounding but a sparse circuit of 2.5-4.8% attention heads in later layers mediates prior-knowledge overrides, identified causally via patching and ablation across three model families.
AirGroundBench is a new diagnostic benchmark exposing that MLLMs handle basic spatial perception but struggle with cross-view alignment, transformation reasoning, and embodied navigation under heterogeneous air-ground views.
VLA language backbones show high redundancy on manipulation benchmarks, with half the LLM blocks removable and even two blocks sufficient to recover baseline performance after fine-tuning, unlike vision and action pathways.
Look-Before-Move is a framework that converts narrative intent into Semantic Observation Contracts, uses Monte Carlo Viewpoint Search for feasible viewpoints, and applies Semantic Trajectory Grounding for coherent camera motion in dynamic 3D story worlds.
citing papers explorer
-
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).
-
MI-CXR: A Benchmark for Longitudinal Reasoning over Multi-Interval Chest X-rays
MI-CXR is a new benchmark that shows state-of-the-art vision-language models achieve only 29.3% accuracy on longitudinal reasoning tasks across multi-visit chest X-ray sequences.
-
Can VLMs Truly Forget? Benchmarking Training-Free Visual Concept Unlearning
VLM-UnBench demonstrates that prompt-based training-free unlearning in VLMs leaves forget accuracy near the no-instruction baseline except under oracle conditions that reveal the target concept.
-
Beyond Pixel Diffs: Benchmarking Image Change Captioning for Web UI Visual Regression Testing
Proposes WUICC task and WUICC-bench dataset, then evaluates 11 image difference captioning methods plus 2 LLMs on web UI changes.
-
WikiVQABench: A Knowledge-Grounded Visual Question Answering Benchmark from Wikipedia and Wikidata
WikiVQABench is a human-curated collection of Wikipedia-based VQA items that require both visual evidence and external knowledge from Wikidata to answer correctly.
-
HalluCXR: Benchmarking and Mitigating Hallucinations in Medical Vision-Language Models for Chest Radiograph Interpretation
HalluCXR benchmark shows 61.9-82.3% hallucination rates across VLMs on MIMIC-CXR images, identifies patterns such as length-based risk and over-fabrication of common findings, and demonstrates ensemble mitigation that cuts fabrication by up to 84.8%.
-
Structured Layout Priors for Robust Out-of-Distribution Visual Document Understanding
Injecting pre-computed layout priors from RT-DETR into VLM prompts raises markdown F1 from 0.37 to 0.92 on a 10k-page OOD benchmark and cuts infinite-loop failures across domains.
-
TOC-Bench: A Temporal Object Consistency Benchmark for Video Large Language Models
TOC-Bench is a new diagnostic benchmark that reveals major weaknesses in temporal object consistency for Video-LLMs, including event counting, ordering, identity reasoning, and hallucination avoidance.
-
One Token Per Frame: Reconsidering Visual Bandwidth in World Models for VLA Policy
Reducing visual input to one token per frame in VLA world models maintains or improves long-horizon performance on MetaWorld, LIBERO, and real-robot tasks.
-
CXR-ContraBench: Benchmarking Negated-Option Attraction in Medical VLMs
Medical VLMs frequently select negated options that contradict visible chest X-ray findings, achieving only ~30% accuracy on direct presence probes, but a post-hoc consistency verifier raises accuracy above 95%.
-
Watching Movies Like a Human: Egocentric Emotion Understanding for Embodied Companions
Creates the first egocentric screen-view movie emotion benchmark and demonstrates that cinematic models drop sharply in Macro-F1 on realistic robot-like viewing conditions while domain-specific training improves robustness.
-
Gaslight, Gatekeep, V1-V3: Early Visual Cortex Alignment Shields Vision-Language Models from Sycophantic Manipulation
Alignment of vision-language models with human V1-V3 early visual cortex negatively predicts resistance to sycophantic gaslighting attacks.
-
UniGeoSeg: Towards Unified Open-World Segmentation for Geospatial Scenes
UniGeoSeg releases the first million-scale dataset for instruction-driven remote sensing segmentation and a unified model that achieves state-of-the-art results with strong zero-shot generalization.
-
High-Resolution Visual Reasoning via Multi-Turn Grounding-Based Reinforcement Learning
MGPO elicits grounding in LMMs via multi-turn RL with binary rewards, yielding 5.4% and 5.2% gains on MME-Realworld and V* Bench and surpassing GPT-4o on the latter after training on 21K samples.
-
Video-Holmes: Can MLLM Think Like Holmes for Complex Video Reasoning?
Video-Holmes benchmark shows top MLLMs achieve at most 45% accuracy on tasks needing integration of multiple clues from suspense films, unlike existing perception-focused tests.
-
Consensus Entropy: Harnessing Multi-VLM Agreement for Self-Verifying and Self-Improving OCR
Consensus Entropy measures inter-VLM output agreement to verify OCR reliability and enable self-improving ensembles, yielding 42.1% F1 gains over single-model judging.
-
Consistent Yet Wrong: Evidence Insensitivity in Spatial Vision-Language Models
Leading VLMs show high cross-view consistency paired with low metric accuracy on distance queries, indicating evidence-insensitive reasoning rather than geometric grounding.
-
Focus-then-Context: Subject-Centric Progressive Visual Token Reduction for Vision-Language Models
SPpruner reduces visual tokens in VLMs via focus identification followed by context-aware scanning, retaining 22.2% tokens for 2.53x speedup on Qwen2.5-VL with negligible accuracy loss.
-
G$^2$TR: Generation-Guided Visual Token Reduction for Separate-Encoder Unified Multimodal Models
G²TR reduces visual tokens and prefill compute by 1.94x in separate-encoder UMMs via generation-guided importance from VAE latent consistency, balanced selection, and merging, while preserving reasoning accuracy and editing quality.
-
SLQ: Bridging Modalities via Shared Latent Queries for Retrieval with Frozen MLLMs
SLQ adapts frozen MLLMs for multimodal retrieval by appending shared latent queries to text and image tokens and introduces KARR-Bench to test knowledge-aware reasoning retrieval.
-
Tango: Taming Visual Signals for Efficient Video Large Language Models
Tango improves video token pruning with diversity-aware selection and ST-RoPE, retaining 98.9% performance on LLaVA-OV at 10% tokens and 1.88x speedup.
-
VERTIGO: Visual Preference Optimization for Cinematic Camera Trajectory Generation
VERTIGO post-trains camera trajectory generators with visual preference signals from Unity-rendered previews scored by a cinematically fine-tuned VLM, cutting character off-screen rates from 38% to near zero while improving framing and prompt adherence.
-
LinMU: Multimodal Understanding Made Linear
LinMU achieves linear-complexity multimodal understanding by swapping self-attention for an M-MATE dual-branch block and distilling from a frozen teacher VLM, matching accuracy with up to 2.7x faster TTFT and 9x higher throughput.
-
SigLino: Efficient Multi-Teacher Distillation for Agglomerative Vision Foundation Models
SigLino distills SigLIP2 and DINOv3 into efficient vision models via asymmetric relation-knowledge distillation, token-balanced batching, and hierarchical data sampling on a new 200M-image corpus, yielding better transfer to grounding VLMs than training from scratch.
-
MedGRPO: Multi-Task Reinforcement Learning for Heterogeneous Medical Video Understanding
MedGRPO applies cross-dataset reward normalization and a clinical LLM judge within multi-task RL to improve vision-language models on heterogeneous medical video understanding tasks using the new MedVidBench dataset.
-
OmniZip: Audio-Guided Dynamic Token Compression for Fast Omnimodal Large Language Models
OmniZip introduces an audio-guided dynamic token compression framework that achieves 3.42X inference speedup and 1.4X memory reduction for omnimodal LLMs without any training.
-
GRAPE: Let GRPO Supervise Query Rewriting by Ranking for Retrieval
GRAPE applies GRPO to an LLM query rewriter with a corpus-relative ranking reward to improve frozen CLIP retrieval by an average 4.9% Recall@10 on shifted benchmarks without retraining or re-embedding.
-
Mitigating Object Hallucinations via Sentence-Level Early Intervention
SENTINEL reduces MLLM object hallucinations by over 90% via sentence-level early intervention with detector-bootstrapped preference data and C-DPO loss, outperforming prior SOTA on hallucination and capability benchmarks.
-
FutureSightDrive: Thinking Visually with Spatio-Temporal CoT for Autonomous Driving
FSDrive uses a generated future scene frame as visual spatio-temporal CoT to improve VLA models for safer autonomous driving trajectory prediction.
-
InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models
InternVL3-78B sets a new open-source SOTA of 72.2 on MMMU via native joint multimodal pre-training, V2PE, MPO, and test-time scaling while remaining competitive with proprietary models.
-
OpenVLThinker: Complex Vision-Language Reasoning via Iterative SFT-RL Cycles
Iterative SFT-RL cycles enable a 7B LVLM to develop sophisticated visual chain-of-thought reasoning and improve performance on math and general reasoning benchmarks.
-
Visual-RFT: Visual Reinforcement Fine-Tuning
Visual-RFT applies reinforcement learning with verifiable perception rewards to improve large vision-language models on fine-grained classification, few-shot detection, and grounding tasks.
-
Retrievals Can Be Detrimental: Unveiling the Backdoor Vulnerability of Retrieval-Augmented Diffusion Models
BadRDM is a backdoor attack on retrieval-augmented diffusion models that poisons the retrieval database with toxicity surrogates and uses multimodal contrastive learning to force toxic generations from text triggers while preserving benign performance.
-
Senna: Bridging Large Vision-Language Models and End-to-End Autonomous Driving
Senna decouples language-based high-level planning from an LVLM with low-level trajectory prediction from an E2E model, reporting 27% lower planning error and 33% lower collisions after pre-training on DriveX and fine-tuning on nuScenes.