而這正是明略科技最新研究成果《PRE-MAP: Personalized Reinforced Eye-tracking Multimodal LLM for High-Resolution Multi-Attribute Point Prediction》的未來價(jià)值所在。目前，該論文以其理論創(chuàng)新和應(yīng)用價(jià)值，已被全球多媒體技術(shù)領(lǐng)域的旗艦會(huì)議ACM MM 2025正式接收，并有望為廣告測量的智能化發(fā)展樹立新范式。
PRE-MAP在訓(xùn)練過程中引入了一種基于強(qiáng)化學(xué)習(xí)的優(yōu)化機(jī)制——Consistency Group Relative Policy Optimization（C-GRPO），通過對預(yù)測點(diǎn)的空間一致性與格式規(guī)范進(jìn)行策略約束，進(jìn)一步增強(qiáng)個(gè)性化注視點(diǎn)預(yù)測的可控性與精度。
Graphical user interfaces (GUIs) are the primary medium for human-computer interaction, yet automating GUI interactions remains challenging due to the complexity of visual elements, dynamic environments, and the need for multi-step reasoning. Existing methods based on vision-language models (VLMs) often suffer
from limited resolution, domain mismatch, and insufficient sequential decisionmaking capability. To address these issues, we propose Mano, a robust GUI agent built upon a multi-modal foundation model pre-trained on extensive web and computer system data. Our approach integrates a novel simulated environment for high-fidelity data generation, a three-stage training pipeline (supervised fine-tuning, offline reinforcement learning, and online reinforcement learning), and a verification module for error recovery. Mano demonstrates state-of-the-art
performance on multiple GUI benchmarks, including Mind2Web and OSWorld, achieving significant improvements in success rate and operational accuracy. Our work provides new insights into the effective integration of reinforcement learning with VLMs for practical GUI agent deployment, highlighting the importance of
domain-specific data, iterative training, and holistic reward design.
數(shù)字廣告測量及驗(yàn)證
一站式社媒營銷監(jiān)測與洞察平臺
報(bào)告.jpg)
微信截圖_590ff29e-e2ac-4e70-8a7e-696ca2b04755.png)
