ABACUS: Adapting Unified Foundation Model for Bridging Image Count Understanding and Generation
Summary
ABACUS is a unified vision-language model that handles multiple counting tasks and count-faithful image generation without benchmark-specific training, achieving state-of-the-art results across seven benchmarks.
View Cached Full Text
Cached at: 06/27/26, 03:48 AM
Paper page - ABACUS: Adapting Unified Foundation Model for Bridging Image Count Understanding and Generation
Source: https://huggingface.co/papers/2606.23835 ABACUS is a unified vision-language model that handles object counting, crowd counting, referring-expression counting, and count-faithful image generation without any benchmark-specific training required. Our model is built on existing 3B-parameter unified foundation model and is adapted for object localization tasks using three key innovations: density-aware adaptive zooming with objectness maps for spatial grounding; a boundary-aware count policy via GRPO to eliminate crop-boundary errors; and a cycle-consistent GRPO strategy where the understanding branch self-critiques generated outputs, closing the understanding-generation gap without any external annotations. ABACUS achieves state-of-the-art results across seven benchmarks, outperforming both task-specific specialists and larger generalist models.
Similar Articles
Count Anything (2 minute read)
Count Anything is a generalist model for text-guided object counting that unifies multiple domains, supported by the new CLOC dataset with 220K images across six visual domains. It achieves strong accuracy and multi-domain generalization.
Count Anything
Count Anything is a generalist vision model for text-guided object counting across multiple domains, using dual-granularity instance enumeration and complementary counting fusion. It achieves strong accuracy and cross-domain generalization, outperforming existing open-world counting methods.
AFUN: Towards an Affordance Foundation Model for Functionality Understanding
AFUN proposes an affordance foundation model that predicts functional masks and 3D motion curves from RGB-D observations and language descriptions, enabling generalizable robot manipulation across diverse environments. The model outperforms baselines on multiple benchmarks and can be deployed for real-world tasks without fine-tuning.
Unified Multimodal Autoregressive Modeling with Shared Context-Visual Tokenizer is Key to Unification
UniAR presents a unified autoregressive framework that uses a single discrete visual tokenizer to bridge visual understanding and generation, achieving state-of-the-art results in image generation and editing.
Covering Human Action Space for Computer Use: Data Synthesis and Benchmark
This paper introduces CUActSpot, a multimodal benchmark for evaluating computer-use agents, and a renderer-based data synthesis pipeline. The proposed Phi-Ground-Any-4B model outperforms open-source models under 32B parameters.