日本久久网I久久精品五月天I91wumaI国产又粗又猛又爽I98福利在线I久操人妻I午夜激情AVI中文字幕在线日亚洲9I色综合久久夜色精品国产天堂I午夜爱爱网站I黄色小说免费观看I精品天堂I亚洲欧美性爱I成人av在线影院I丝袜网站黄I91色国产I毛片基地,无码I姝姝窝人体色www聚色窝

全球雙榜SOTA！明略科技專有大模型 Mano開(kāi)啟GUI智能操作新時(shí)代

2025-09-28

明略科技推出的GUI大模型Mano在Mind2Web和OSWorld兩大基準(zhǔn)測(cè)試中取得了創(chuàng)紀(jì)錄的SOTA成績(jī)，成功率達(dá)到40.1%。通過(guò)在線強(qiáng)化學(xué)習(xí)和自動(dòng)數(shù)據(jù)采集，Mano為GUI智能體領(lǐng)域提供了可擴(kuò)展的新范式，顯著提升了復(fù)雜任務(wù)的執(zhí)行能力。該技術(shù)的突破不僅推動(dòng)了自動(dòng)化邊界的擴(kuò)展，也為企業(yè)智能化轉(zhuǎn)型提供了強(qiáng)有力的支持

官宣！明略科技推出專有大模型產(chǎn)品線DeepMiner，可信商業(yè)數(shù)據(jù)分析智能體終于能用了！

2025-09-22

明略科技推出專有大模型產(chǎn)品線DeepMiner，以“可信智能體+可信數(shù)據(jù)”雙輪驅(qū)動(dòng)，解決企業(yè)智能體落地中幻覺(jué)率高、過(guò)程不透明等痛點(diǎn)。該產(chǎn)品通過(guò)多智能體協(xié)同架構(gòu)、企業(yè)級(jí)數(shù)據(jù)整合及全流程透明化設(shè)計(jì)，降低幻覺(jué)率并支持知識(shí)沉淀，其自研Mano和Cito模型分別實(shí)現(xiàn)精準(zhǔn)執(zhí)行與深度推理，助力企業(yè)構(gòu)建可信生產(chǎn)力。

明略科技 Mano Technical Report

2025-09-18

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.