INDEX

Explanations

negatively connoted words or phrases related to disapproval and concealment

oai_token-act-pair · gpt-4o-mini Triggered by @bot

ending in "appro", "conce", "scri", "pri" or "ple"

np_acts-logits-general · gemini-2.0-flash

disapproving, conceiving, describing, prescribing, depriving

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 22-gemmascope-res-16k

Configuration

google/gemma-scope-2b-pt-res/layer_22/width_16k/average_l0_72

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.22.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 tabPage

-0.50

 Numer

-0.49

onets

-0.49

 Quint

-0.48

 FileWriter

-0.48

 blister

-0.47

ριο

-0.47

 Brau

-0.47

dk

-0.46

 Sandi

-0.46

POSITIVE LOGITS

ConstraintMaker

0.68

 незавершена

0.63

EDEFAULT

0.63

клопе

0.60

 TestBed

0.59

 unknownFields

0.57

AsUp

0.55

новниш

0.55

Према

0.54

->___

0.53

Activations Density 0.355%

negatively connoted words or phrases related to disapproval and concealment

ending in "appro", "conce", "scri", "pri" or "ple"

disapproving, conceiving, describing, prescribing, depriving

No Comments

No Known Activations