x

LESSWRONG

LW

Pushpita Das

Pushpita Das

Message

Hi! I'm Pushpita. I'm a computational astrophysicist turning to AI safety.

I did my PhD at the University of Amsterdam and then spent some time as a postdoctoral fellow at Columbia University (New York) working on numerical simulations of black holes. My research background has mostly...

9

1

4mo

Pushpita Das

Hi! I'm Pushpita. I'm a computational astrophysicist turning to AI safety.

I did my PhD at the University of Amsterdam and then spent some time as a postdoctoral fellow at Columbia University (New York) working on numerical simulations of black holes. My research background has mostly...

Pushpita Das — LessWrong

The Geometry of Yes: Mapping Sycophancy Inside an LLM's Emotion Space

Summary LLMs have internal emotion representations that causally shape their behaviour. This was recently observed in Claude: positive emotions like happy and loving are linked to sycophancy, and you can steer the model by directly manipulating these directions in activation space. But some questions were left open. Here we focus...