EVE enables verifiable self-evolution of MLLMs by using a Challenger-Solver architecture to generate dynamic executable visual transformations that produce VQA problems with absolute execution-verified ground truth.
hub Mixed citations
Qwen3-VL-Embedding and Qwen3-VL-Reranker: A Unified Framework for State-of-the-Art Multimodal Retrieval and Ranking
Mixed citation behavior. Most common role is background (43%).
31
Pith papers citing it
Background
43% of classified citations
abstract
In this report, we introduce the Qwen3-VL-Embedding and Qwen3-VL-Reranker model series, the latest extensions of the Qwen family built on the Qwen3-VL foundation model. Together, they provide an end-to-end pipeline for high-precision multimodal search by mapping diverse modalities, including text, images, document images, and video, into a unified representation space. The Qwen3-VL-Embedding model employs a multi-stage training paradigm, progressing from large-scale contrastive pre-training to reranking model distillation, to generate semantically rich high-dimensional vectors. It supports Matryoshka Representation Learning, enabling flexible embedding dimensions, and handles inputs up to 32k tokens. Complementing this, Qwen3-VL-Reranker performs fine-grained relevance estimation for query-document pairs using a cross-encoder architecture with cross-attention mechanisms. Both model series inherit the multilingual capabilities of Qwen3-VL, supporting more than 30 languages, and are released in $\textbf{2B}$ and $\textbf{8B}$ parameter sizes to accommodate diverse deployment requirements. Empirical evaluations demonstrate that the Qwen3-VL-Embedding series achieves state-of-the-art results across diverse multimodal embedding evaluation benchmarks. Specifically, Qwen3-VL-Embedding-8B attains an overall score of $\textbf{77.8}$ on MMEB-V2, ranking first among all models (as of January 8, 2025). This report presents the architecture, training methodology, and practical capabilities of the series, demonstrating their effectiveness on various multimodal retrieval tasks, including image-text retrieval, visual question answering, and video-text matching.
hub tools
citation-role summary
background 6
baseline 4
method 3
dataset 1
citation-polarity summary
years
2026 31representative citing papers
FashionMV introduces product-level multi-view CIR, a 127K-product dataset built via automated LMM pipeline, and a 0.8B ProCIR model that beats larger baselines on three fashion benchmarks.
ReTool-Video uses a 134-tool meta-augmented library and recursive grounding to translate abstract video intents into fine-grained multimodal operations, outperforming baselines on MVBench, MLVU, and Video-MME.
FLARE is a new benchmark with 399 long videos, 87k multimodal clips, and 275k user-style queries for testing audiovisual retrieval under caption and query regimes.
Omni-Persona benchmark with 18 tasks shows open-source models have audio-visual grounding gaps, RLVR narrows them but leads to conservative outputs, and scale or recall alone fail as diagnostics.
MMEB-V3 benchmark shows omni-modality embedding models fail to enforce instruction-specified modality constraints and exhibit asymmetric, query-biased retrieval.
CLAY reframes pretrained VLM embedding spaces as text-conditional similarity spaces for adaptive, multi-conditioned image retrieval without additional training.
VISAFF is a tuning-free speaker-centered visual affective feature learning framework for emotion recognition in conversation that guides frozen VLMs to active speakers and uses reliability-guided complementation from textual and acoustic modalities to achieve competitive performance.
TIGER-FG proposes text-guided implicit fine-grained grounding with dual distillation to address modality and granularity asymmetries in image-to-multimodal e-commerce retrieval, reporting Recall@1 gains of 6.1 and 34.4 points on two new benchmarks.
GELATO extends frozen text embedding models with locked image and audio encoders, training minimal connectors to produce a single semantic embedding space for text, image, audio, and video while keeping original text performance unchanged.
MINER fuses internal transformer layer representations via probing and adaptive sparse fusion to improve dense single-vector retrieval quality on visual documents by up to 4.5% nDCG@5 while preserving efficiency.
Uncertainty estimation for LLM hallucinations can be done effectively with partial generations or input-only predictors, reducing the need for full autoregressive sampling.
A single-pass black-box method models LLM outputs as dynamical systems via Koopman operators to detect hallucinations with claimed state-of-the-art accuracy and lower cost.
ScrapMem introduces optical forgetting to compress multimodal memories for LLM agents on edge devices, cutting storage by up to 93% while reaching 51.0% Joint@10 and 70.3% Recall@10 on ATM-Bench.
A scalable training-free pipeline using video segmentation, filtering, and off-the-shelf multimodal models creates DenseStep2M, a dataset of 100K videos and 2M detailed instructional steps that improves dense captioning, step grounding, and cross-modal retrieval.
Patch Forcing enables diffusion models to denoise image patches at varying rates based on predicted difficulty, advancing easier regions first to improve context and achieve better generation quality on ImageNet while scaling to text-to-image tasks.
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.
MemJack achieves 71.48% attack success rate on unmodified COCO val2017 images against Qwen3-VL-Plus by coordinating agents to map visual entities to malicious intents, apply multi-angle camouflage, and filter refusals via iterative nullspace projection while transferring strategies through a shared
A vision-language-aligned world model turns visuomotor MPC into a language-following planner that reaches 87% success on 288 unseen semantic tasks where standard VLAs drop to 22%.
Tempo uses a 6B SVLM as a local temporal compressor with training-free adaptive token allocation to achieve SOTA long-video understanding at 0.5-16 tokens per frame, scoring 52.3 on 4101s LVBench under 8K budget.
Semantic-level UI Element Injection distracts GUI agents by overlaying safety-aligned UI elements, achieving up to 4.4x higher attack success rates that transfer across models and create persistent attractors.
HIVE raises multimodal retrieval nDCG@10 to 41.7 on the MM-BRIGHT benchmark by inserting LLM-driven hypothesis generation and verification between retrieval passes, delivering +9.5 over the best text-only baseline and +14.1 over the best multimodal baseline.
MMEmb-R1 adaptively applies chain-of-thought reasoning to multimodal embeddings via pair-aware counterfactual selection and RL, reaching 71.2 on MMEB-V2 with a 4B model and lower latency.
LucidNFT combines a new LR-referenced consistency reward, decoupled normalization, and a real-degradation dataset to improve perceptual quality in flow-matching super-resolution while preserving input fidelity.
citing papers explorer
-
EVE: Verifiable Self-Evolution of MLLMs via Executable Visual Transformations
EVE enables verifiable self-evolution of MLLMs by using a Challenger-Solver architecture to generate dynamic executable visual transformations that produce VQA problems with absolute execution-verified ground truth.
-
FashionMV: Product-Level Composed Image Retrieval with Multi-View Fashion Data
FashionMV introduces product-level multi-view CIR, a 127K-product dataset built via automated LMM pipeline, and a 0.8B ProCIR model that beats larger baselines on three fashion benchmarks.
-
ReTool-Video: Recursive Tool-Using Video Agents with Meta-Augmented Tool Grounding
ReTool-Video uses a 134-tool meta-augmented library and recursive grounding to translate abstract video intents into fine-grained multimodal operations, outperforming baselines on MVBench, MLVU, and Video-MME.
-
FLARE: Full-Modality Long-Video Audiovisual Retrieval Benchmark with User-Simulated Queries
FLARE is a new benchmark with 399 long videos, 87k multimodal clips, and 275k user-style queries for testing audiovisual retrieval under caption and query regimes.
-
Omni-Persona: Systematic Benchmarking and Improving Omnimodal Personalization
Omni-Persona benchmark with 18 tasks shows open-source models have audio-visual grounding gaps, RLVR narrows them but leads to conservative outputs, and scale or recall alone fail as diagnostics.
-
MMEB-V3: Measuring the Performance Gaps of Omni-Modality Embedding Models
MMEB-V3 benchmark shows omni-modality embedding models fail to enforce instruction-specified modality constraints and exhibit asymmetric, query-biased retrieval.
-
CLAY: Conditional Visual Similarity Modulation in Vision-Language Embedding Space
CLAY reframes pretrained VLM embedding spaces as text-conditional similarity spaces for adaptive, multi-conditioned image retrieval without additional training.
-
VISAFF: Speaker-Centered Visual Affective Feature Learning for Emotion Recognition in Conversation
VISAFF is a tuning-free speaker-centered visual affective feature learning framework for emotion recognition in conversation that guides frozen VLMs to active speakers and uses reliability-guided complementation from textual and acoustic modalities to achieve competitive performance.
-
TIGER-FG: Text-Guided Implicit Fine-Grained Grounding for E-commerce Retrieval
TIGER-FG proposes text-guided implicit fine-grained grounding with dual distillation to address modality and granularity asymmetries in image-to-multimodal e-commerce retrieval, reporting Recall@1 gains of 6.1 and 34.4 points on two new benchmarks.
-
jina-embeddings-v5-omni: Geometry-preserving Embeddings via Locked Aligned Towers
GELATO extends frozen text embedding models with locked image and audio encoders, training minimal connectors to produce a single semantic embedding space for text, image, audio, and video while keeping original text performance unchanged.
-
MINER: Mining Multimodal Internal Representation for Efficient Retrieval
MINER fuses internal transformer layer representations via probing and adaptive sparse fusion to improve dense single-vector retrieval quality on visual documents by up to 4.5% nDCG@5 while preserving efficiency.
-
Towards Generation-Efficient Uncertainty Estimation in Large Language Models
Uncertainty estimation for LLM hallucinations can be done effectively with partial generations or input-only predictors, reducing the need for full autoregressive sampling.
-
Low-Cost Black-Box Detection of LLM Hallucinations via Dynamical System Prediction
A single-pass black-box method models LLM outputs as dynamical systems via Koopman operators to detect hallucinations with claimed state-of-the-art accuracy and lower cost.
-
ScrapMem: A Bio-inspired Framework for On-device Personalized Agent Memory via Optical Forgetting
ScrapMem introduces optical forgetting to compress multimodal memories for LLM agents on edge devices, cutting storage by up to 93% while reaching 51.0% Joint@10 and 70.3% Recall@10 on ATM-Bench.
-
DenseStep2M: A Scalable, Training-Free Pipeline for Dense Instructional Video Annotation
A scalable training-free pipeline using video segmentation, filtering, and off-the-shelf multimodal models creates DenseStep2M, a dataset of 100K videos and 2M detailed instructional steps that improves dense captioning, step grounding, and cross-modal retrieval.
-
Denoising, Fast and Slow: Difficulty-Aware Adaptive Sampling for Image Generation
Patch Forcing enables diffusion models to denoise image patches at varying rates based on predicted difficulty, advancing easier regions first to improve context and achieve better generation quality on ImageNet while scaling to text-to-image tasks.
-
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.
-
Every Picture Tells a Dangerous Story: Memory-Augmented Multi-Agent Jailbreak Attacks on VLMs
MemJack achieves 71.48% attack success rate on unmodified COCO val2017 images against Qwen3-VL-Plus by coordinating agents to map visual entities to malicious intents, apply multi-angle camouflage, and filter refusals via iterative nullspace projection while transferring strategies through a shared
-
Grounded World Model for Semantically Generalizable Planning
A vision-language-aligned world model turns visuomotor MPC into a language-following planner that reaches 87% success on 288 unseen semantic tasks where standard VLAs drop to 22%.
-
Small Vision-Language Models are Smart Compressors for Long Video Understanding
Tempo uses a 6B SVLM as a local temporal compressor with training-free adaptive token allocation to achieve SOTA long-video understanding at 0.5-16 tokens per frame, scoring 52.3 on 4101s LVBench under 8K budget.
-
Are GUI Agents Focused Enough? Automated Distraction via Semantic-level UI Element Injection
Semantic-level UI Element Injection distracts GUI agents by overlaying safety-aligned UI elements, achieving up to 4.4x higher attack success rates that transfer across models and create persistent attractors.
-
HIVE: Query, Hypothesize, Verify An LLM Framework for Multimodal Reasoning-Intensive Retrieval
HIVE raises multimodal retrieval nDCG@10 to 41.7 on the MM-BRIGHT benchmark by inserting LLM-driven hypothesis generation and verification between retrieval passes, delivering +9.5 over the best text-only baseline and +14.1 over the best multimodal baseline.
-
MMEmb-R1: Reasoning-Enhanced Multimodal Embedding with Pair-Aware Selection and Adaptive Control
MMEmb-R1 adaptively applies chain-of-thought reasoning to multimodal embeddings via pair-aware counterfactual selection and RL, reaching 71.2 on MMEB-V2 with a 4B model and lower latency.
-
LucidNFT: LR-Anchored Multi-Reward Preference Optimization for Flow-Based Real-World Super-Resolution
LucidNFT combines a new LR-referenced consistency reward, decoupled normalization, and a real-degradation dataset to improve perceptual quality in flow-matching super-resolution while preserving input fidelity.
-
CausalEmbed: Auto-Regressive Multi-Vector Generation in Latent Space for Visual Document Embedding
CausalEmbed uses auto-regressive generation with iterative margin loss to produce multi-vector embeddings that reduce visual token counts 30-155x while retaining competitive performance on VDR benchmarks.
-
A Picture is Worth a Thousand Words? An Empirical Study of Aggregation Strategies for Visual Financial Document Retrieval
Single-vector aggregation in visual financial document retrieval collapses semantically distinct documents due to global texture dominance, as demonstrated by a new diagnostic benchmark where patch-level signals detect changes that aggregated vectors obscure.
-
MicroWorld: Empowering Multimodal Large Language Models to Bridge the Microscopic Domain Gap with Multimodal Attribute Graph
MicroWorld constructs a multimodal attributed property graph from scientific image-caption data and augments MLLM prompts via retrieval to raise Qwen3-VL-8B performance by 37.5% on MicroVQA and 6% on MicroBench.
-
VL-SAM-v3: Memory-Guided Visual Priors for Open-World Object Detection
VL-SAM-v3 retrieves visual prototypes from memory to generate sparse spatial and dense contextual priors that refine detection prompts, yielding gains on rare categories in LVIS for both open-vocabulary and open-ended settings.
-
Enhancing Multimodal In-Context Learning via Inductive-Deductive Reasoning
A framework with similarity-based visual token compression, dynamic attention rebalancing, and explicit inductive-deductive chain-of-thought improves multimodal ICL performance across eight benchmarks for open-source VLMs.
-
OpenSpatial: A Principled Data Engine for Empowering Spatial Intelligence
OpenSpatial supplies a principled open-source data engine and 3-million-sample dataset that raises spatial-reasoning model performance by an average of 19 percent on benchmarks.
-
Neuroscience-Inspired Analyses of Visual Interestingness in Multimodal Transformers
Human visual interestingness is linearly decodable from final-layer embeddings in Qwen3-VL-8B and becomes progressively more structured across vision and language layers without explicit supervision.