A deep dive into Potemkin understanding in LLMs, why current benchmarks miss it, and why it matters for measuring what models really understand.
Oh no LLMs are acting like stochastic next word predictors!
Hahah. Although I wouldn't underestimate their emergent capabilities.
The problem seems to be rather wild overestimation/anthropomorphisation though.
Is there any way I can read or start from basics regarding LLM?
You’ll have to start with the basics of ML if you’re new to it. Directly learning about LLMs is difficult.
Oh no LLMs are acting like stochastic next word predictors!
Hahah. Although I wouldn't underestimate their emergent capabilities.
The problem seems to be rather wild overestimation/anthropomorphisation though.
Is there any way I can read or start from basics regarding LLM?
You’ll have to start with the basics of ML if you’re new to it. Directly learning about LLMs is difficult.