INDEX

Explanations

imperative verbs or calls to action

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

instructional verbs

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_131k/average_l0_81

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.20.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

 ComVisible

-0.50

 <=",

-0.48

StructEnd

-0.47

 ControllerBase

-0.45

Aholisi

-0.45

GEBURTSDATUM

-0.43

"])

-0.42

'))

-0.42

bcryptjs

-0.42

Efq

-0.41

POSITIVE LOGITS

colgante

0.54

 noDo

0.53

pantalón

0.49

raise

0.49

spesies

0.49

Apply

0.49

 uppdater

0.47

Rispondi

0.47

lappe

0.47

DoNot

0.47

Activations Density 0.073%

imperative verbs or calls to action

instructional verbs

No Comments

No Known Activations