Introduces PexelsCustom-1M dataset, CustoMDiT parameter-efficient model, and OpenCustom benchmark for open-domain customized video generation.
hub
Florence-2: Advancing a unified representation for a variety of vision tasks (2023)
15 Pith papers cite this work, alongside 14 external citations. Polarity classification is still indexing.
15
Pith papers citing it
14
external citations · Pith
abstract
We introduce Florence-2, a novel vision foundation model with a unified, prompt-based representation for a variety of computer vision and vision-language tasks. While existing large vision models excel in transfer learning, they struggle to perform a diversity of tasks with simple instructions, a capability that implies handling the complexity of various spatial hierarchy and semantic granularity. Florence-2 was designed to take text-prompt as task instructions and generate desirable results in text forms, whether it be captioning, object detection, grounding or segmentation. This multi-task learning setup demands large-scale, high-quality annotated data. To this end, we co-developed FLD-5B that consists of 5.4 billion comprehensive visual annotations on 126 million images, using an iterative strategy of automated image annotation and model refinement. We adopted a sequence-to-sequence structure to train Florence-2 to perform versatile and comprehensive vision tasks. Extensive evaluations on numerous tasks demonstrated Florence-2 to be a strong vision foundation model contender with unprecedented zero-shot and fine-tuning capabilities.
hub tools
citation-role summary
background 1
citation-polarity summary
roles
background 1polarities
background 1representative citing papers
AnchorSeg uses ordered query banks of latent reasoning tokens plus a spatial anchor token and a Token-Mask Cycle Consistency loss to achieve 67.7% gIoU and 68.1% cIoU on the ReasonSeg benchmark.
ASTRA disentangles subject identity from pose structure in diffusion transformers via retrieval-augmented pose guidance, asymmetric EURoPE embeddings, and a DSM adapter to improve multi-subject generation.
DBAC is a new directional metric for bias amplification in image captions that is less sensitive to sentence encoders and more accurate than LIC, validated on COCO gender and race attributes.
A single RGB image is converted into a layered, simulation-ready robot scene that supports trajectory replay, synthetic data generation, and meaningful sim-real policy evaluation, plus a 564-scene DROID-Sim companion set.
SADL introduces the first real-world benchmark for subject-aware distractor localization with 1,800 cases and evaluates VLMs showing they over-exclude objects.
T-IMPACT is a new benchmark dataset and pipeline that supplies nearly 99k manipulated image-text pairs together with a human-calibrated continuous severity signal for contextual interpretation change.
Unlabeled egocentric video can be converted into obstacle-aware goal-conditioned navigation supervision through monocular geometry, ESDFs, and MPPI-planned trajectories.
ERVLA trains on a 978k-trajectory embodied CoT corpus using reasoning as supervision with dropout, then predicts actions without CoT at test time, reaching 86.9% on LIBERO-Plus and 53.2% on VLABench.
BVE framework enables text-guided 3D editing beyond voxel limits by combining self-constructed data, lightweight semantic injection, and annotation-free masking to preserve local invariance.
MINGLE is a modular pipeline that combines off-the-shelf detection tools with VLM reasoning to localize socially connected groups in urban scenes and is supported by a new 100K-image dataset.
Weak-to-strong knowledge distillation applied early and then turned off accelerates convergence to target performance in visual learning tasks by factors of 1.7-4.8x.
Qwen2.5-VL reports a vision-language model family using native dynamic-resolution ViT and absolute time encoding that matches GPT-4o on document and diagram tasks while supporting hour-long videos with second-level localization.
PaliGemma is an open 3B VLM based on SigLIP and Gemma that achieves strong performance on nearly 40 diverse open-world tasks including benchmarks, remote-sensing, and segmentation.
PointArena 2026 uses agent-driven data synthesis, cross-block attention, and iterative correction to achieve 77.2% accuracy on a visual pointing benchmark for embodied AI.
citing papers explorer
-
A Comprehensive Ecosystem for Open-Domain Customized Video Generation
Introduces PexelsCustom-1M dataset, CustoMDiT parameter-efficient model, and OpenCustom benchmark for open-domain customized video generation.
-
AnchorSeg: Language Grounded Query Banks for Reasoning Segmentation
AnchorSeg uses ordered query banks of latent reasoning tokens plus a spatial anchor token and a Token-Mask Cycle Consistency loss to achieve 67.7% gIoU and 68.1% cIoU on the ReasonSeg benchmark.
-
ASTRA: Enhancing Multi-Subject Generation with Retrieval-Augmented Pose Guidance and Disentangled Position Embedding
ASTRA disentangles subject identity from pose structure in diffusion transformers via retrieval-augmented pose guidance, asymmetric EURoPE embeddings, and a DSM adapter to improve multi-subject generation.
-
A Woman with a Knife or A Knife with a Woman? Measuring Directional Bias Amplification in Image Captions
DBAC is a new directional metric for bias amplification in image captions that is less sensitive to sentence encoders and more accurate than LIC, validated on COCO gender and race attributes.
-
RoboSnap: One-Shot Real-to-Sim Scene Generation for Generalizable Robot Learning and Evaluation
A single RGB image is converted into a layered, simulation-ready robot scene that supports trajectory replay, synthetic data generation, and meaningful sim-real policy evaluation, plus a 564-scene DROID-Sim companion set.
-
SADL: What to Ignore? A Benchmark for Subject-Aware Distractor Localization
SADL introduces the first real-world benchmark for subject-aware distractor localization with 1,800 cases and evaluates VLMs showing they over-exclude objects.
-
T-IMPACT: A Severity-Aware Benchmark for Contextual Image-Text Manipulation
T-IMPACT is a new benchmark dataset and pipeline that supplies nearly 99k manipulated image-text pairs together with a human-calibrated continuous severity signal for contextual interpretation change.
-
VEGA: Learning Navigation VLAs from In-the-Wild Egocentric Video with Geometric Trajectory Supervision
Unlabeled egocentric video can be converted into obstacle-aware goal-conditioned navigation supervision through monocular geometry, ESDFs, and MPPI-planned trajectories.
-
Revisiting Embodied Chain-of-Thought for Generalizable Robot Manipulation
ERVLA trains on a 978k-trajectory embodied CoT corpus using reasoning as supervision with dropout, then predicts actions without CoT at test time, reaching 86.9% on LIBERO-Plus and 53.2% on VLABench.
-
Beyond Voxel 3D Editing: Learning from 3D Masks and Self-Constructed Data
BVE framework enables text-guided 3D editing beyond voxel limits by combining self-constructed data, lightweight semantic injection, and annotation-free masking to preserve local invariance.
-
MINGLE: VLMs for Semantically Complex Region Detection in Urban Scenes
MINGLE is a modular pipeline that combines off-the-shelf detection tools with VLM reasoning to localize socially connected groups in urban scenes and is supported by a new 100K-image dataset.
-
Weak-to-Strong Knowledge Distillation Accelerates Visual Learning
Weak-to-strong knowledge distillation applied early and then turned off accelerates convergence to target performance in visual learning tasks by factors of 1.7-4.8x.
-
Qwen2.5-VL Technical Report
Qwen2.5-VL reports a vision-language model family using native dynamic-resolution ViT and absolute time encoding that matches GPT-4o on document and diagram tasks while supporting hour-long videos with second-level localization.
-
PaliGemma: A versatile 3B VLM for transfer
PaliGemma is an open 3B VLM based on SigLIP and Gemma that achieves strong performance on nearly 40 diverse open-world tasks including benchmarks, remote-sensing, and segmentation.
-
Efficient Visual Pointing for Embodied AI:Agent-Driven Data Synthesis, Cross-Block Attention, and Iterative Correction
PointArena 2026 uses agent-driven data synthesis, cross-block attention, and iterative correction to achieve 77.2% accuracy on a visual pointing benchmark for embodied AI.