INDEX

Explanations

code `#include`, `namespace`, `class`

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

preprocessor directives and file includes in source code (especially `#include` statements and blank lines following them).

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

The neuron fires most strongly on C/C++ preprocessor directives (e.g. the “#” token and include-guard identifiers like “render_…h”) and related constructs.

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

 then

-2.59

 нынеш

-1.47

eping

-1.41



-1.38

 latter

-1.38

然後

-1.38

 advise

-1.35

 этом

-1.34

 agência

-1.31

 call

-1.30

POSITIVE LOGITS

 через

1.55

 если

1.47

 neuf

1.41

 утеп

1.38

 Япо

1.34

剧照

1.32

сторо

1.31

 merusak

1.31

 اگر

1.30

 звез

1.30

Activations Density 0.004%

code `#include`, `namespace`, `class`

preprocessor directives and file includes in source code (especially `#include` statements and blank lines following them).

The neuron fires most strongly on C/C++ preprocessor directives (e.g. the “#” token and include-guard identifiers like “render_…h”) and related constructs.

No Comments

No Known Activations

code `#include`, `namespace`, `class`

preprocessor directives and file includes in source code (especially `#include` statements and blank lines following them).

The neuron fires most strongly on C/C++ preprocessor directives (e.g. the “#” token and include-guard identifiers like “render_…h”) and related constructs.

No Comments

No Known Activations