INDEX

Explanations

drug or technical names

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

proper nouns or name-like terms, particularly those with unusual syllable combinations.

oai_token-act-pair · claude-3-7-sonnet-20250219

Triggered by @neilrathi

This neuron strongly activates on two‐letter character bigrams (especially “ro,” “HE,” “KA,” “TT,” etc.) wherever they appear within tokens.

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

Negative Logits

.”

-1.90

for

-1.79

our

-1.70

 impropri

-1.66

hline

-1.56

。

-1.54

Symptom

-1.54

 which

-1.51

 controladores

-1.51

 الأولى

-1.50

POSITIVE LOGITS

is

2.06

 evangé

1.89

谵

1.84

 recentes

1.79

It

1.76

కర్య

1.73

镚

1.70

awarkan

1.70

If

1.69

 cercano

1.69

Activations Density 0.135%

drug or technical names

proper nouns or name-like terms, particularly those with unusual syllable combinations.

This neuron strongly activates on two‐letter character bigrams (especially “ro,” “HE,” “KA,” “TT,” etc.) wherever they appear within tokens.

No Known Activations

drug or technical names

proper nouns or name-like terms, particularly those with unusual syllable combinations.

This neuron strongly activates on two‐letter character bigrams (especially “ro,” “HE,” “KA,” “TT,” etc.) wherever they appear within tokens.

No Known Activations