INDEX

Explanations

file type or method

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

Chinese and other non-English (particularly East Asian and Cyrillic script) text.

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

The neuron fires on content‐words that name a “way” or “means” (e.g. method, format, tool, weapon).

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

🪬

-2.83

 parachoque

-2.67

甞

-2.64

眎

-2.61

.";

-2.58

蟊

-2.52

 cittadino

-2.48

.」

-2.42

 westfalen

-2.39

 fiore

-2.39

POSITIVE LOGITS

ка

2.59

You

2.34

 einem

2.31

 Many

2.25

There

2.25

меня

2.19

 aktuellen

2.17

行く

2.16

 erase

2.14

 Presumably

2.13

Activations Density 0.000%

file type or method

Chinese and other non-English (particularly East Asian and Cyrillic script) text.

The neuron fires on content‐words that name a “way” or “means” (e.g. method, format, tool, weapon).

No Comments

No Known Activations

file type or method

Chinese and other non-English (particularly East Asian and Cyrillic script) text.

The neuron fires on content‐words that name a “way” or “means” (e.g. method, format, tool, weapon).

No Comments

No Known Activations