INDEX

Explanations

computer networking, USS Hamlin, hypothalamic

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

legal and technical terms like "habeas corpus", "homotopy", and "ubiquitin".

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

Triggered by @neilrathi

The neuron strongly activates on longer, infrequent “jargon” tokens—specialized or technical roots (e.g. habeas, homotopy, ubiquit, interpretation, camel-cased marginLeft).

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

If

-3.06

AND

-2.95

 如果

-2.64

 When

-2.61

も

-2.48

 Если

-2.41

 This

-2.34

 este

-2.33

And

-2.30

 Even

-2.30

POSITIVE LOGITS

畱

2.69

ﱠ

2.67

只是一个

2.56

已经在

2.55

皝

2.53

跸

2.52

神

2.41

ᯅ

2.39

韮

2.38

襌

2.38

Activations Density 0.001%

computer networking, USS Hamlin, hypothalamic

legal and technical terms like "habeas corpus", "homotopy", and "ubiquitin".

The neuron strongly activates on longer, infrequent “jargon” tokens—specialized or technical roots (e.g. habeas, homotopy, ubiquit, interpretation, camel-cased marginLeft).

No Known Activations

computer networking, USS Hamlin, hypothalamic

legal and technical terms like "habeas corpus", "homotopy", and "ubiquitin".

The neuron strongly activates on longer, infrequent “jargon” tokens—specialized or technical roots (e.g. habeas, homotopy, ubiquit, interpretation, camel-cased marginLeft).

No Known Activations