Zhijing Jin

An­gles of at­tack for con­tinual learn­ing safety

How might con­tinual learn­ing af­fect safety and al­ign­ment?

What’s Con­tinual Learn­ing, and Why Might We Ex­pect To See It In Ad­vanced LLM Agents?

Im­pli­ca­tions of Con­tinual Learn­ing for LLM Agents: Introduction

Ex­plain­ing un­de­sir­able model be­hav­ior: (How) can in­fluence func­tions help?

The Multi-Agent Minefield: Can LLMs Co­op­er­ate to Avoid Global Catas­tro­phe?

Test­ing the Author­i­tar­ian Bias of LLMs

Why Rea­son­ing Isn’t Enough: How LLM Agents Strug­gle with Ethics and Cooperation

In­ves­ti­gat­ing Ac­ci­den­tal Misal­ign­ment: Causal Effects of Fine-Tun­ing Data on Model Vulnerability

Cor­rupted by Rea­son­ing: Rea­son­ing Lan­guage Models Be­come Free-Riders in Public Goods Games

• Zhijing Jin 27 Sep 2023 16:10 UTC

  1 point

  0

  in reply to: Mo Putera’s comment on: Wel­come to Ap­ply: The 2024 Vi­talik Bu­terin Fel­low­ships in AI Ex­is­ten­tial Safety by FLI!

  Thank you for spotting it! I just did the fix :).

Wel­come to Ap­ply: The 2024 Vi­talik Bu­terin Fel­low­ships in AI Ex­is­ten­tial Safety by FLI!