INDEX

Explanations

convert data types

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

the word "convert" or adverbs ending in "-ly" (like "purely", "practically").

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

The neuron fires on the first substantive content word in a segment (i.e. it skips past function words like “is,” “a,” “the,” etc., and activates on the first non–stopword).

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

感じ

-2.25

↵

-2.22

 dejamos

-2.19

 größe

-2.16

皤

-2.06

 grün

-2.05

</i>

-2.05

 addition

-1.99

-1.97

 нескольких

-1.95

POSITIVE LOGITS

en

3.67

as

2.97

曩

2.84

 zusätzliche

2.52

呖

2.47

螃

2.34

 zahlreiche

2.33

Ｅ

2.30

er

2.27

”？

2.27

Activations Density 0.000%

convert data types

the word "convert" or adverbs ending in "-ly" (like "purely", "practically").

The neuron fires on the first substantive content word in a segment (i.e. it skips past function words like “is,” “a,” “the,” etc., and activates on the first non–stopword).

No Comments

No Known Activations

convert data types

the word "convert" or adverbs ending in "-ly" (like "purely", "practically").

The neuron fires on the first substantive content word in a segment (i.e. it skips past function words like “is,” “a,” “the,” etc., and activates on the first non–stopword).

No Comments

No Known Activations