INDEX

Explanations

action calls after prompts

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

phrases expressing ability or capability, particularly "be able to" and "can" followed by action verbs.

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

This neuron fires on imperative action verbs that describe user interactions (e.g. click, press, enter, save, go, choose, check).

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

愛用

-1.34

Дина

-1.23

ciutto

-1.20

 enduring

-1.18

 increased

-1.16

 Umgang

-1.14

INCRE

-1.13

 continued

-1.13

↵↵

-1.12

 found

-1.12

POSITIVE LOGITS

and

2.08

 zonder

1.57

 that

1.51

 fön

1.31

')}

1.28

 takimi

1.27

 quella

1.27

значение

1.25

身形

1.24

 filtres

1.23

Activations Density 0.140%

action calls after prompts

phrases expressing ability or capability, particularly "be able to" and "can" followed by action verbs.

This neuron fires on imperative action verbs that describe user interactions (e.g. click, press, enter, save, go, choose, check).

No Comments

No Known Activations

action calls after prompts

phrases expressing ability or capability, particularly "be able to" and "can" followed by action verbs.

This neuron fires on imperative action verbs that describe user interactions (e.g. click, press, enter, save, go, choose, check).

No Comments

No Known Activations