INDEX

Explanations

parentheses in code

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

closing parentheses that end names, titles, or abbreviations in formal or technical text.

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

The neuron strongly activates on capitalized proper names and titles (e.g. the headings or article titles at the start of entries).

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

lossenen

-1.20

>",

-1.18

]=>

-1.17

%");

-1.16

 "}";

-1.16

нительно

-1.15

}");

-1.13

()/

-1.12

kasten

-1.10

studie

-1.10

POSITIVE LOGITS

 несмотря

1.22

 после

1.22

for

1.20

 vínculos

1.15

as

1.09

шее

1.08

 Normally

1.06

 destacado

1.06

 Initially

1.06

 después

1.06

Activations Density 0.271%

parentheses in code

closing parentheses that end names, titles, or abbreviations in formal or technical text.

The neuron strongly activates on capitalized proper names and titles (e.g. the headings or article titles at the start of entries).

No Comments

No Known Activations

parentheses in code

closing parentheses that end names, titles, or abbreviations in formal or technical text.

The neuron strongly activates on capitalized proper names and titles (e.g. the headings or article titles at the start of entries).

No Comments

No Known Activations