INDEX

Explanations

Genetic and drug possession

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

programming syntax elements in code files, particularly variable and file naming patterns with underscores.

oai_token-act-pair · claude-3-7-sonnet-20250219 Triggered by @neilrathi

The neuron activates most strongly on programming identifiers (or tensor names) that include trailing numeric indices (e.g. “conv2d_1”, “input_230”, “init_127.5”).

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

New Auto-Interp

Configuration

google/gemma-scope-27b-pt-res/layer_22/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

#%%

-1.44

只剩

-1.14

-1.11

 conditions

-1.11

 Livro

-1.10

So

-1.08

一路

-1.06

 Then

-1.05

The

-1.05

下班

-1.05

POSITIVE LOGITS

 reager

1.48

Daha

1.45

 tarjeta

1.44

猻

1.38

饷

1.37

Tienes

1.32

 personaje

1.29

 Muhammadu

1.27

 maneiras

1.26

Jumlah

1.23

Activations Density 0.004%

Genetic and drug possession

programming syntax elements in code files, particularly variable and file naming patterns with underscores.

The neuron activates most strongly on programming identifiers (or tensor names) that include trailing numeric indices (e.g. “conv2d_1”, “input_230”, “init_127.5”).

No Comments

No Known Activations

Genetic and drug possession

programming syntax elements in code files, particularly variable and file naming patterns with underscores.

The neuron activates most strongly on programming identifiers (or tensor names) that include trailing numeric indices (e.g. “conv2d_1”, “input_230”, “init_127.5”).

No Comments

No Known Activations