INDEX

Explanations

commands and actions

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

camelCase function or method names in code (particularly those with "And" in the middle).

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

The neuron fires on occurrences of standalone action-words—especially verbs used as commands or function/operation names (e.g. Go, return, take, charge, skip).

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

 that

-1.79

you

-1.76

 både

-1.34

''

-1.33

 nawet

-1.32

↵↵↵

-1.28

有任何

-1.27

jenige

-1.26

jenigen

-1.25

这个人

-1.25

POSITIVE LOGITS

 verres

1.33

 oreilles

1.31

for

1.31

ꪔ

1.29

熥

1.23

 carreaux

1.22

 épaules

1.20

was

1.20

 jaunes

1.20

lösung

1.16

Activations Density 0.064%

commands and actions

camelCase function or method names in code (particularly those with "And" in the middle).

The neuron fires on occurrences of standalone action-words—especially verbs used as commands or function/operation names (e.g. Go, return, take, charge, skip).

No Comments

No Known Activations

commands and actions

camelCase function or method names in code (particularly those with "And" in the middle).

The neuron fires on occurrences of standalone action-words—especially verbs used as commands or function/operation names (e.g. Go, return, take, charge, skip).

No Comments

No Known Activations