NOW LET US – AI RAG SaaS Studio TP.HCM
NOW LET US
Digital Product Studio
Back to news
AGENTIC-SYSTEMS...1 min read

COMPASS: Grounding Composition-Intent Guidance in Unified Multimodal Models

Share
NOW LET US Article – COMPASS: Grounding Composition-Intent Guidance in Unified Multimodal Models

Researchers introduce COMPASS, the first unified multimodal framework that bridges composition perception and controlled generation using a shared expert token. Supported by the new Comp-11 dataset, COMPASS significantly improves AI's ability to understand and generate images with precise layout control.

Computer Science > Artificial Intelligence

Title:COMPASS: Grounding Composition-Intent Guidance in Unified Multimodal Models

View PDF HTML (experimental)Abstract:Composition is a high-level visual intent that governs where subjects are placed and how a scene is organized, yet current unified multimodal models remain unreliable at fine-grained composition recognition and struggle to turn such intent into controllable generation. We present COMPASS, the first unified multimodal framework that grounds composition-intent control in a single system spanning both composition perception and composition-guided generation, with a shared expert token $\tau_c$ as the central intent anchor. On the perception side, COMPASS injects composition expertise into an MoE backbone in a minimally invasive manner and distills the inferred intent into $\tau_c$. On the generation side, COMPASS reuses $\tau_c$ as a global conditioning signal that steers the denoising trajectory, effectively converting passive composition analysis into explicit layout control. To support systematic instruction-following composition learning and evaluation at scale, we construct Comp-11, a large-scale dataset with an 11-class taxonomy and reasoning-augmented annotations. Extensive experiments show that COMPASS substantially improves category-level composition understanding and delivers more composition-consistent, prompt-faithful generation than strong baselines.

Bibliographic and Citation Tools

Code, Data and Media Associated with this Article

Demos

Recommenders and Search Tools

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

© 2026 Now Let Us. All rights reserved.

Source: arXiv cs.AI Recent

Advertisement
Ad slot ready: 5887729102

More in this category

NOW LET US Related – Primary ICD Category Prediction using LLM-based Probing

agentic-systems

Primary ICD Category Prediction using LLM-based Probing

Researchers evaluated whether frozen medical large language model (LLM) representations can serve as a shared embedding space for multimodal primary diagnosis category prediction, achieving high accuracy and efficient cross-dataset transfer.

NOW LET US Related – ComMem: Complementary Memory Systems for Test-Time Adaptation of Vision-Language Models

agentic-systems

ComMem: Complementary Memory Systems for Test-Time Adaptation of Vision-Language Models

Researchers have proposed ComMem, an innovative test-time adaptation (TTA) approach for vision-language models that mimics the complementary memory systems of the biological brain to enhance practical adaptability.

NOW LET US Related – TrajRS: Towards Certified Robustness in Pedestrian Trajectory Prediction

agentic-systems

TrajRS: Towards Certified Robustness in Pedestrian Trajectory Prediction

Researchers have introduced TrajRS, a new framework that enhances and certifies the robustness of pedestrian trajectory prediction models against adversarial attacks, marking a significant step toward safer autonomous driving systems.

NOW LET US Related – Search for Truth from Reasoning: A Dynamic Representation Editing Framework for Steering LLM Trajectories

agentic-systems

Search for Truth from Reasoning: A Dynamic Representation Editing Framework for Steering LLM Trajectories

Researchers have introduced DynaSteer, a dynamic representation editing framework designed to guide Large Language Models (LLMs) toward truthful reasoning trajectories. By dynamically intervening at high-entropy decision points, DynaSteer improves reasoning accuracy without disrupting correct logical paths.

NOW LET US Related – An AI agent for treatment reasoning over a biomedical tool universe

agentic-systems

An AI agent for treatment reasoning over a biomedical tool universe

Researchers have introduced ATHENA-R1, a breakthrough AI agent capable of advanced treatment reasoning across all FDA-approved drugs since 1939. Trained via reinforcement learning over 212 biomedical tools, it outperforms leading models like GPT-5 in complex medical decision-making.

NOW LET US Related – BV-Blend: Uncertainty-Weighted Historical Baselines for Stable Critic-Free RL with Verifiable Rewards

agentic-systems

BV-Blend: Uncertainty-Weighted Historical Baselines for Stable Critic-Free RL with Verifiable Rewards

Researchers have introduced BV-Blend, a critic-free reinforcement learning framework that stabilizes advantage estimation in LLM alignment. By blending historical data with local statistics, BV-Blend overcomes a critical limitation of GRPO-style methods during cold-start training with binary rewards.

EXPLORE TOPICS

Discover All Categories

Deep dive into the specific technology sectors that matter most to you.

AI Frontier

AI Frontier

Foundation models, multimodal systems, and enterprise use cases.

Agentic Systems

Agentic Systems

Task-oriented automation workflows for real operations.

SaaS Growth

SaaS Growth

CAC, LTV, retention, and growth optimization.

Dev Tools

Dev Tools

Codegen tools, next-gen IDEs, and modern dev workflows.

Cybersecurity

Cybersecurity

AI security, Zero Trust, and proactive defense.

Cloud & Infra

Cloud & Infra

Cloud Native, Edge Computing, and GPU Clusters.

Robotics

Robotics

Humanoid robots, autonomous vehicles, and smart hardware.

Startups & VC

Startups & VC

Funding trends, M&A, and the startup ecosystem.