微软让 GPT 与 Claude 协同工作，性能超越所有竞品
ChainCatcher 消息，微软周一宣布为 Copilot Researcher 推出两项新功能—Critique 与 Council，将 OpenAI 的 GPT 与 Anthropic 的 Claude 结合用于同一研究任务。
Critique 采用串联协作模式：GPT 负责规划研究、检索资料并生成初稿，Claude 随后担任严格审阅者，核查事实准确性与引用质量；Council 则让两个模型并行独立生成报告，再由第三个裁判模型对比差异、归纳分歧。
在涵盖医疗、法律、科技等 10 个领域共 100 项复杂研究任务的 DRACO 基准测试中，搭载 Critique 的 Copilot 得分 57.4 分，领先第二名近 14%，远超 Claude Opus 4.6 单独运行的 42.7 分。
Disclaimer: OKX Orbit content is provided for informational purposes only. Learn more
Replies
Related Flash News
Quantum computing company Quantinuum is seeking to raise $1.05 billion through an IPO
Wintermute: Bitcoin's key support level is in the $75,000–$76,000 range, indicating that market structure has not fully deteriorated
Iranian media: Unfreezing Iranian funds is the last serious obstacle between Iran and the United States
Bloomberg: 9 whale wallets dominate Polymarket's multi-billion dollar dispute ruling
The U.S. Department of Defense and SpaceX have disputed Starlink pricing
BSTR's Chief Investment Officer stated that they are building "Berkshire Hathaway 2," aiming to increase the number of Bitcoins per share
The path of least resistance for gold prices remains downward
Iran insists that half of the frozen funds will be available when the agreement is announced
Analysis: Bitcoin has cooled significantly in short-term turnover, and the market is near the bottom
Iranian media denied claims that Iran and the U.S. had reached a memorandum of understanding
