为什么？ 因为通用聊天机器人的竞争已经卷到头了。 真正能赚钱的战场，是金融研报、科研综述、投研咨询这些高价值场景。 新智元报道 编辑：定慧 Aeneas 【新智元导读】百度版Deep Research（百度千帆深度研究Agent）横空出世，首秀即拿下权威榜单TOP 1。当AI真正学会「做研究」，金融、科研等高净 值行业的游戏规则，全球千亿美元市场正在被彻底改写。 从2025到2026， Deep Research成了AI巨头们的兵家必争之地。 OpenAI押上了，谷歌押上了，Anthropic也押上了。 而这些场景需要的，不是能聊天，是能把复杂问题研究明白。 这正是Deep Research要解决的问题！ 就在这个节骨眼上，百度出手了。 2月4日，深度研究智能体权威评测榜单DeepResearch Bench公布最新结果，百度千帆深度研究Agent直接登顶榜首，一举超越OpenAI、Gemini、Claude等 一众国际顶尖选手的同类产品。 | | | | | | | | | The research ams to comprehensively evaluate the capabilities of De ...

Core Insights - The article discusses the evolution of Deep Research (DR) as a new direction in AI, moving from simple dialogue and creative writing applications to more complex research-oriented tasks. It highlights the limitations of traditional retrieval-augmented generation (RAG) methods and introduces DR as a solution for multi-step reasoning and long-term research processes [2][30]. Summary by Sections Definition of Deep Research - DR is not a specific model or technology but a progressive capability pathway for research-oriented agents, evolving from information retrieval to complete research workflows [5]. Stages of Capability Development - **Stage 1: Agentic Search** - Models gain the ability to actively search and retrieve information dynamically based on intermediate results, focusing on efficient information acquisition [5]. - **Stage 2: Integrated Research** - Models evolve to understand, filter, and integrate multi-source evidence, producing coherent reports [6]. - **Stage 3: Full-stack AI Scientist** - Models can propose research hypotheses, design and execute experiments, and reflect on results, emphasizing depth of reasoning and autonomy [6]. Core Components of Deep Research - **Query Planning** - Involves deciding what information to query next, incorporating dynamic adjustments in multi-round research [10]. - **Information Retrieval** - Focuses on when to retrieve, what to retrieve, and how to filter retrieved information to avoid redundancy and ensure relevance [12][13][14]. - **Memory Management** - Essential for long-term reasoning, involving memory consolidation, indexing, updating, and forgetting [15]. - **Answer Generation** - Stresses the logical consistency between conclusions and evidence, requiring integration of multi-source evidence [17]. Training and Optimization Methods - **Prompt Engineering** - Involves designing multi-step prompts to guide the model through research processes, though its effectiveness is highly dependent on prompt design [20]. - **Supervised Fine-tuning** - Utilizes high-quality reasoning trajectories for model training, though acquiring annotated data can be costly [21]. - **Reinforcement Learning for Agents** - Directly optimizes decision-making strategies in multi-step processes without complex annotations [22]. Challenges in Deep Research - **Coordination of Internal and External Knowledge** - Balancing reliance on internal reasoning versus external information retrieval is crucial [24]. - **Stability of Training Algorithms** - Long-term task training often faces issues like policy degradation, limiting exploration of diverse reasoning paths [24]. - **Evaluation Methodology** - Developing reliable evaluation methods for research-oriented agents remains an open question, with existing benchmarks needing further exploration [25][27]. - **Memory Module Construction** - Balancing memory capacity, retrieval efficiency, and information reliability is a significant challenge [28]. Conclusion - Deep Research represents a shift from single-turn answer generation to in-depth research addressing open-ended questions. The field is still in its early stages, with ongoing exploration needed to create autonomous and trustworthy DR agents [30].

AI Agents & Enterprise Solutions - Box introduces AI agents for deep research, general search, and data extraction across enterprise data [5] - AI agents enable tasks like due diligence on company acquisitions by analyzing thousands of documents, previously too expensive or impractical [6][7] - Box's AI agents aim to increase productivity by automating tasks that were previously not undertaken due to cost or time constraints [10] Competitive Advantage & Differentiation - Box differentiates itself as an enterprise platform with over 120,000 customers, including 64% of the Fortune 500, emphasizing data security and management [11] - Box allows enterprises to bring AI to their content, rather than moving data to different AI platforms [12] - Box integrates across the AI ecosystem, working with models from Gemini, OpenAI, Meta, XAI, Grok, Claude, and Anthropic [13] - Box integrates into the IT stack of enterprise customers with IBM Watson X Orchestrate, Salesforce Agent Force, and Google Agent Space [13] Market Impact & Customer Adoption - AI agents are expected to be used primarily for tasks that were previously not done by humans due to cost or complexity [9][10] - Box has engaged with over 100 customers in Q1, with most deploying AI in use cases where humans were not previously involved [9]

Financial Data and Key Metrics Changes - Total company organic revenues rose by 7%, with the big three segments growing by 9% [4][10] - Adjusted EBITDA increased by 10% to $672 million, reflecting a margin increase to 37.7% [10][26] - Adjusted EPS was $0.85 for the quarter, compared to $0.80 in the prior year [26] Business Line Data and Key Metrics Changes - Legal Professionals segment saw organic revenue growth of 9%, up from 8% in the first half of 2025 [6][11] - Corporates segment organic revenues grew by 7%, driven by offerings in Legal, Tax, and Risk portfolios [11][23] - Tax and Accounting organic revenues grew by 10%, supported by strong performance in Latin America and the U.S. [11][24] - Reuters News organic revenues rose by 3%, primarily due to growth in the agency business [12][25] - Global Print organic revenues declined by 4% year on year [12][25] Market Data and Key Metrics Changes - The percentage of annualized contract value from GenAI-enabled products increased to 24%, up from 22% in the previous quarter [25] - The company expects organic revenue growth of approximately 7% in Q4, including about 9% for the big three segments [31] Company Strategy and Development Direction - The company is focused on leveraging AI innovations to enhance product offerings, particularly in the Legal Professionals and Tax and Accounting segments [5][21] - A balanced capital allocation approach is maintained, with a commitment to assess additional inorganic opportunities [8][9] - The company completed a $1 billion share repurchase program and remains well-capitalized with a net leverage of only 0.6 times [8][9] Management's Comments on Operating Environment and Future Outlook - Management reaffirmed the full-year 2025 revenue and profit outlook, expecting approximately 9% organic revenue growth for the big three segments [4][29] - Temporary factors affecting growth include slower commercial print volumes, U.S. Federal government cancellations, and softer bookings trends in Corporates [5][29] - The company is optimistic about the long-term value proposition in government despite recent downgrades and cancellations [44][45] Other Important Information - The company is updating its 2026 financial framework, expecting organic revenue growth of 7.5% to 8% and margin expansion [32] - Free cash flow outlook for 2026 is raised to approximately $2.1 billion, reflecting confidence in operational efficiency [33] Q&A Session Summary Question: What are the recurring impacts of government and corporate headwinds? - Management remains confident in the corporate segment's growth potential despite temporary sales softness, expecting 9% to 11% organic growth next year [40][43] Question: How is customer reaction to AgenTik AI? - Customer feedback has been very positive, with significant changes in behavior noted among users, indicating strong adoption of the new tools [45][46] Question: What is the impact of the government shutdown on contracts? - Cancellations occurred prior to the shutdown, which has minimal impact on current revenue [52] Question: How does the company view competition in the AI assistant space? - New players have entered the AI assistant space, but the company is confident in its position and product development plans [68] Question: What is the pricing strategy in light of AI product value? - The company follows a price-to-value principle, ensuring pricing aligns with the efficiencies provided by AI products [93][94] Question: How does the company see AI affecting the tax business? - The tax business is expected to benefit from AI advancements, enabling more efficient processes and advisory services [114][115]

@Tether_to @GoKiteAI Addendum 2: @caesar_data adds x402 payment supportAgents can now make micropayments to access one of the best deep research models availableThis is gonna snowball real quick; AI szn imminenthttps://t.co/y3jwyPmf7D https://t.co/1Iwlpn8XimCaesar (@caesar_data):Caesar x402 support is now live.Developers and AI agents can now access the best deep research on demand, using trustless, instant pay-per-query payments powered by the x402 protocol. https://t.co/V4StlUeZ82 ...

Product Assessment - The product "Caesar" is considered strong after testing [1] - The product is recommended as a Deep Research tool, especially for crypto analysis/research [1] Team & Venture - The venture is backed by experience and data engineering expertise from the team [1]

Product Assessment - Caesar is a strong deep research tool, particularly for crypto analysis/research [1] - The product's strength has impressed early testers [1] Team & Expertise - The team possesses wealth of experience and data engineering expertise [1] - Backing the venture was a logical decision due to the team's expertise [1]

Product Assessment - The product "Caesar" is considered strong after testing [1] - The product is recommended as a Deep Research tool, especially for crypto analysis/research [1] Team & Venture - The venture is backed by experience and data engineering expertise from the team [1]

Core Insights - MiroMind aims to build a self-aware digital agent ecosystem, focusing on the continuous evolution of Artificial General Intelligence (AGI) through community collaboration and open-source principles [2][4]. Group 1: Open Source Ecosystem - MiroMind has developed a comprehensive open-source ecosystem that includes the Agent framework (MiroFlow), models (MiroThinker), data (MiroVerse), and training infrastructure (MiroTrain/MiroRL), all of which are open for learning, reuse, and further development [1][8]. - The MiroFlow framework achieved a state-of-the-art (SOTA) score of 82.4 on the GAIA validation set, surpassing existing commercial model APIs [1][12]. - MiroThinker, the core model, reached a SOTA performance of 60.2% on the GAIA-Text-103 dataset, nearing the performance level of OpenAI's Deep Research [1][15]. Group 2: Community Collaboration - MiroMind fosters a developer-centric environment that encourages community participation through data requests, feature customization, and technical challenges, with feedback directly influencing project development [2][22]. - The project organizes various community activities such as competitions, leaderboards, and hackathons to enhance developer engagement and contribution [22]. Group 3: Key Personnel - The project is led by Chen Tianqiao, a renowned entrepreneur known for his strategic vision and significant contributions to brain science and AI [4]. - Dai Jifeng, a key figure in the project, is a professor at Tsinghua University with extensive experience in computer vision and deep learning, having published over 80 papers with significant citations [5][6].

Core Viewpoint - The AI sector, particularly the "AI Four Strong" companies, is experiencing a resurgence as they pivot towards Deep Research and AI Agents to compete with larger firms and demonstrate their value to investors [1][3][21]. Group 1: AI Four Strong's Strategy - The AI Four Strong have launched their own Deep Research products, aiming to regain market presence after a period of low activity [2][9]. - These companies are focusing on vertical integration and delivering value through Deep Research and AI Agents, which are seen as safer positions amid competition from larger firms [3][4]. - The introduction of AI Agents is not only a strategy to re-enter the market but also a way to create monetizable opportunities, with reports indicating significant user upgrades to premium services [5][21]. Group 2: Market Dynamics and Competition - The AI landscape has shifted with the emergence of Deep Research as a new benchmark, prompting the AI Four Strong to innovate rapidly [8][17]. - The competition has intensified, with major players like Tencent and Alibaba also entering the fray, leading to a reassessment of strategies among the AI Four Strong [17][20]. - The AI Four Strong are now prioritizing technological advancements over user growth, reflecting a strategic shift in response to market pressures [20][21]. Group 3: Product Development and User Engagement - The AI Four Strong are adopting different approaches to product development, with some focusing on user-friendly interfaces while others emphasize high user interaction [12][13]. - Recent model releases, such as MiniMax's M1 and Kimi's K2, showcase significant advancements in capabilities, including increased parameter counts and improved efficiency [15][23]. - The need for AI Agents to deliver quantifiable value to clients is critical, as demonstrated by successful case studies that highlight efficiency improvements and cost reductions [27][28]. Group 4: Financial Outlook and Investment - The AI Four Strong are beginning to attract positive attention from investors, with reports of significant funding rounds and IPO plans [23][24]. - The financial viability of AI Agents is under scrutiny, as the costs associated with their use can be substantial, necessitating a focus on creating clear value propositions for clients [30][31]. - The overall market for AI Agents is still developing, with indications that achieving product-market fit remains a challenge for many companies in this space [31][32].