INDEX

Explanations

VALUES, creds,imgs

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

the word "values" or "vals" in programming code contexts.

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

This neuron fires on identifier tokens that are plural common variable names ending in “s,” such as “creds,” “values,” and “maxs.”

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

玙

-2.97

-2.83

鉅

-2.56

饪

-2.56

钜

-2.47



-2.45

㌘

-2.41

しようと

-2.36

睥

-2.36

嗩

-2.31

POSITIVE LOGITS

“

3.52

2.72

<strong>

2.47

我

2.45

er

2.45

「

2.39

”:

2.33

<em>

2.33

↵↵

2.31

Ꮟ

2.31

Activations Density 0.003%

VALUES, creds,imgs

the word "values" or "vals" in programming code contexts.

This neuron fires on identifier tokens that are plural common variable names ending in “s,” such as “creds,” “values,” and “maxs.”

No Comments

No Known Activations

VALUES, creds,imgs

the word "values" or "vals" in programming code contexts.

This neuron fires on identifier tokens that are plural common variable names ending in “s,” such as “creds,” “values,” and “maxs.”

No Comments

No Known Activations