La dialectique est la vie profonde de cette contradiction, la série des progrès qu'elle accomplit. Une histoire qui se fait et qui cependant est à faire, un sens qui n’est jamais nul, mais toujours à rectifier, à reprendre, à maintenir contre les hasards, un savoir qu’aucun irrationnel positif ne limite, mais qui pourtant ne contient pas actuellement la totalité du réel accompli et à accomplir, et dont le pouvoir d’exhaustion est à prouver par le fait, une histoire-réalité qui est juge ou critère de toutes nos pensées, mais qui elle-même n’est autre chose que l’avènement de la conscience, de sorte que nous n’avons pas à lui obéir passivement, mais à la penser selon nos propres forces.

—— Maurice Merleau-Ponty, Les aventures de la dialectique. (1955)

My research lies at the intersection of Large Language Models and Agentic AI, focusing on how autonomous agents plan, reason, and act in complex, open-ended environments. I am particularly interested in multi-agent systems—how agents coordinate, communicate, and even compete to achieve collective goals, and how useful behaviors emerge from their interactions. Reinforcement learning is a central thread throughout my work, serving as the key lever for enhancing the decision-making and long-horizon planning capabilities of LLM-based agents. This research interest is complemented by hands-on industry experience in LLM post-training—spanning reinforcement learning from human feedback, instruction tuning, and alignment—which grounds my understanding of how these techniques translate from theory into production-scale systems. Ultimately, I aim to build agentic systems that are more capable, reliable, and aligned with human intentions, and I believe that AGI will be achieved through AI bootstrapping and self-evolution.

• AI-Related Philosophy: Human-Centeredness in the Age of Acceleration; the Paradigm Shift Toward Post-Anthropocentrism; the Subjectivity and Intersubjectivity of Agents.
• Indie Game Development: Visual novel, tabletop RPG, and CRPG enthusiast — dreaming of one day making something as great as Disco Elysium and Ever17.
• Quantitative Finance: Leveraging data and models to discover and reliably monetize predictable patterns across markets, including A-shares, Hong Kong equities, NASDAQ, and cryptocurrencies.
• Systems Science & Engineering: Thinking in feedback loops, emergence, and hierarchies — applying system dynamics and cybernetics to architect and steer complex sociotechnical systems where reductionism fails.

千万个你 渺小的无法从此辨别 就像拜科努尔的沙 | 也许泪水不觉间已落下 因为此刻国境线正在眼中融化 | 化作流星 化作晚风 化作朵飘落的花 | 100公里还只是个起点 更远的天空(Небо) 还有许多奇迹要写下

—— “科罗廖夫十字绽开，卡门线上化作花束”【航天史填词】 (2025)