INDEX

Explanations

phrases that indicate definitions or explanations

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

identifying what something is called

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

antMatchers

-0.80

featureID

-0.74

vician

-0.64

 anún

-0.61

 feroit

-0.60

 ciudadana

-0.59

 desierto

-0.57

 desmotivaciones

-0.56

 navideña

-0.55

majánló

-0.55

POSITIVE LOGITS

 called

0.45

 term

0.41

 name

0.41

OGND

0.41

 commonly

0.38

val

0.36

 Called

0.35

WriteTagHelper

0.35

called

0.35

 common

0.33

Activations Density 0.052%

phrases that indicate definitions or explanations

identifying what something is called

No Comments

No Known Activations