INDEX

Explanations

names and technical terms

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

Japanese surnames or names (particularly those containing "Kura" or ending in common Japanese name patterns).

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

The neuron is highly sensitive to subword fragments of proper names and technical terms (e.g. author surnames, cell‐type names, scientific taxonomy, etc.).

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

-2.95

𐐪

-2.48

 parfü

-2.48

是

-2.44

鲠

-2.41



-2.38

 这个

-2.22

紲

-2.19

 laajenta

-2.17

 größte

-2.16

POSITIVE LOGITS

 siendo

2.42

屙

2.34



2.34

 laden

2.22

跺

2.20

ʻ

2.19

 fueron

2.17

犛

2.14



2.13

is

2.11

Activations Density 0.004%

names and technical terms

Japanese surnames or names (particularly those containing "Kura" or ending in common Japanese name patterns).

The neuron is highly sensitive to subword fragments of proper names and technical terms (e.g. author surnames, cell‐type names, scientific taxonomy, etc.).

No Comments

No Known Activations

names and technical terms

Japanese surnames or names (particularly those containing "Kura" or ending in common Japanese name patterns).

The neuron is highly sensitive to subword fragments of proper names and technical terms (e.g. author surnames, cell‐type names, scientific taxonomy, etc.).

No Comments

No Known Activations