INDEX

Explanations

list items and code blocks

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

closing tags in code, particularly HTML/XML closing tags and the ends of string or list items.

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

The neuron fires on human‐readable label text—especially capitalized words used as menu items, link texts, form options, or other UI labels embedded in code.

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

 weiteres

-1.58

 anderes

-1.41

 Otros

-1.31

-1.30

 like

-1.29

 aforementioned

-1.27

the

-1.26

Besitos

-1.22

 Tinggi

-1.17

อื่น

-1.16

POSITIVE LOGITS

 they

1.72

acamole

1.61

zembro

1.55

we

1.44

 befindet

1.43

是什么意思

1.43

这种情况

1.38

 respectable

1.37

{~

1.36

</td>

1.34

Activations Density 0.032%

list items and code blocks

closing tags in code, particularly HTML/XML closing tags and the ends of string or list items.

The neuron fires on human‐readable label text—especially capitalized words used as menu items, link texts, form options, or other UI labels embedded in code.

No Comments

No Known Activations

list items and code blocks

closing tags in code, particularly HTML/XML closing tags and the ends of string or list items.

The neuron fires on human‐readable label text—especially capitalized words used as menu items, link texts, form options, or other UI labels embedded in code.

No Comments

No Known Activations