INDEX

Explanations

difficult

The neuron fires on words and small function phrases that express modality or evaluation—i.e. modals (“can,” “be,” “has”), hedges and qualifiers (“difficult,” “impossible,” “proved,” “more,” “increases”) that signal possibility, ability, or degree.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

ylül

0.59

ayo

0.53

Yay

0.51

పరి

0.49

plr

0.45

ទ្

0.45

 bekend

0.45

ಂತ್ರಿ

0.45

स्म

0.44

vline

0.44

POSITIVE LOGITS

 difficult

3.56

difficult

3.42

Difficult

3.33

 Difficult

3.27

 difícil

3.23

 DIFFIC

3.22

 difficile

3.16

 difíciles

3.09

 трудно

3.05

 moeilijk

3.05

Activations Density 0.592%