INDEX

Explanations

Ek and Ik prefixes

np_acts-logits-general · gemini-2.5-flash-lite

the letter sequence "Ek" appearing in proper nouns (names, places, and usernames).

oai_token-act-pair · claude-4-5-sonnet Triggered by @bfgcn7frf5

This neuron detects two-letter uppercase tokens ending in “k,” such as “Ek” or “Ik.”

oai_token-act-pair · o4-mini Triggered by @jyhe0408

New Auto-Interp

Configuration

google/gemma-scope-27b-pt-res/layer_10/width_131k

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

-2.47

gées

-2.44

<i>

-2.41

י

-2.36

技术的

-2.33

 ,‬

-2.25

bní

-2.23

嬛

-2.20

,…

-2.17

遢

-2.14

POSITIVE LOGITS

?”

2.67

end

2.67

ons

2.55

2.48

 them

2.39

As

2.36

"?

2.31

}\

2.27

redhat

2.27

 neuem

2.23

Activations Density 0.002%

Ek and Ik prefixes

the letter sequence "Ek" appearing in proper nouns (names, places, and usernames).

This neuron detects two-letter uppercase tokens ending in “k,” such as “Ek” or “Ik.”

No Comments

No Known Activations

Ek and Ik prefixes

the letter sequence "Ek" appearing in proper nouns (names, places, and usernames).

This neuron detects two-letter uppercase tokens ending in “k,” such as “Ek” or “Ik.”

No Comments

No Known Activations