MATS 2026 年秋季项目
2026 年秋季项目为期 10 周，于 9 月 28 日至 12 月 4 日在美国加州伯克利和英国伦敦开展。研究员将接受顶尖研究人员以及 Anthropic、Google DeepMind、OpenAI、Redwood Research 和 ARC 等机构研究人员的指导，并可申请在主项目结束后继续参加 6 至 12 个月的带资助延长期。本届首次开设创业与领域建设方向和生物安全方向。
Applications are now closed.
Program phases
Key dates for the application and admissions timeline
Applications typically open several months before the program begins. Applicants complete a multi-stage admissions process, beginning with a general application. Depending on the tracks, streams, and mentors they apply to, applicants may also complete additional evaluations such as interviews, work tests, coding assessments, or writing samples before final admissions decisions are made.
第一阶段，申请人可申请一个或多个研究方向：实证研究、理论、战略与预测、政策与治理、系统安全、生物安全，以及创业与领域建设。进入第二阶段后，申请人可选择这些方向下的具体研究流，每个研究流由一位或多位拥有各自研究议程的导师负责。你可以查看网格视图。
Peter Wildeford
I am hoping to get more help developing policy ideas for future policy windows.
The Redwood Research stream is looking for fast empirical iterators and strategists to work on control research.
Eleos AI Research is a nonprofit working to ensure the interests of AI systems are appropriately taken into account as we navigate transformative AI. We are looking for competent generalists to help scale up our AI welfare field-building by running high-impact events, or to otherwise amplify our research, communications, and operational capacity.
Backing projects focused on product development and organization building in the areas of AI safety and alignment, biosecurity, and critical cybersecurity. Looking for fellows who are self starters, default to action, and have a desire to create.
即使技术持续进步，疗法也可能始终比病原体更具优势。在生物风险情景下，如何让疗法研发与生产更具韧性？又该如何最大限度地利用 AI 进展来加强防御？
This stream will work on projects that empirically assess national security threats of AI misuse (CBRN terrorism and cyberattacks) and improve dangerous capability evaluations. Threat modeling applicants should have a skeptical mindset, enjoy case study work, and be strong written communicators. Eval applicants should be able and excited to help demonstrate concepts like sandbagging elicitation gaps in an AI misuse context.
This stream will focus on impact-oriented technical AI governance research work, potentially including research on open-weight models, applied AI safeguards research, AI incidents, technically rigorous AI policy, etc.
In the shard theory stream, we create qualitatively new methods and fields of inquiry, from steering vectors to gradient routing to unsupervised capability elicitation to robust unlearning. If you're theory-minded, maybe you'll help us formalize shard theory itself.