Steven LIU

NLP, LLM, Agent

AI Researcher, PhD

[Zhihang Fu @Fostor_HUNT]

Our unified framework models five coupled components:

• foundation model
• agent harness
• data system
• trainer
• improvement mechanism

The key RSI step: the system improves not only its capabilities, but also how it improves.

[Zhihang Fu @Fostor_HUNT]

We define self-improvement as persistent updates to an agent system’s own components—not merely adaptation within one run.

We then introduce an L1–L5 capability scale, from manual improvement to general recursive self-improvement.

[Zhihang Fu @Fostor_HUNT]

What does it take for an AI agent to improve not just its behavior, but the process by which it improves?

We’re excited to share our new survey:

The Path to Recursive Self-Improving Agents
Foundation. Framework. Future directions.

Paper: https://t.co/xmXmVQsMcz

[Steven LIU @Steven_Lau6]

Rubber Duck @HK

[Steven LIU @Steven_Lau6]

😅😅😅

[Steven LIU @Steven_Lau6]

😅😅😅

[Steven LIU @Steven_Lau6]

@gdb @sama How about Hong Kong？😁

[Steven LIU @Steven_Lau6]

As Chinese researchers try new AI products released in the past few weeks 😣😣😣

[Steven LIU @Steven_Lau6]

@xiye_nlp 😭😭😭

[Steven LIU @Steven_Lau6]

@xathis @OpenAI Thanks for your great work! I'm interested in which pdf parsing library you used. We tried some open-source libraries for our NLP research before. But their parsing results are not so good, especially when the pdf typesetting is complex.

[Steven LIU @Steven_Lau6]

NYT today😂

[Steven LIU @Steven_Lau6]

👀

[Steven LIU @Steven_Lau6]

Pre-training of Deep Bidirectional Transformers for Car Wash😆

[Steven LIU @Steven_Lau6]

Chicago Marathon😊

[Steven LIU @Steven_Lau6]

@WilliamWangNLP Found the policy here https://t.co/9en0zLmP3q

[Steven LIU @Steven_Lau6]

@WilliamWangNLP Many authors received comments asking to compare with methods in NAACL22 papers. 😂 Is there any guideline for this?

[Steven LIU @Steven_Lau6]

@emnlpmeeting Just confirmed it's not April Fool's Day😆