INDEX

Explanations

code structure identifiers

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

structured data elements in code, particularly class names, object properties, and data structure definitions.

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

The neuron fires on tokens that look like programming‐style identifiers or API/JSON field names (e.g. class or variable names, keys in code or JSON).

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

 announces

-1.15

 perv

-1.14

eve

-1.14

 announcements

-1.13

 lavori

-1.12

 bisog

-1.09

 alimentazione

-1.09

eb

-1.09

 himself

-1.09

 urgent

-1.09

POSITIVE LOGITS

 that

1.29

 parseFrom

1.26

 سپس

1.19

该

1.14

 několik

1.13

 ویتامین

1.12

 gewährleisten

1.09

 courtier

1.09

是可以

1.08

 bandung

1.07

Activations Density 0.046%

code structure identifiers

structured data elements in code, particularly class names, object properties, and data structure definitions.

The neuron fires on tokens that look like programming‐style identifiers or API/JSON field names (e.g. class or variable names, keys in code or JSON).

No Comments

No Known Activations

code structure identifiers

structured data elements in code, particularly class names, object properties, and data structure definitions.

The neuron fires on tokens that look like programming‐style identifiers or API/JSON field names (e.g. class or variable names, keys in code or JSON).

No Comments

No Known Activations