INDEX

Explanations

references to specific parts or components in a technical or procedural text

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

conductive layer

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

leyebilirsiniz

-0.42

IALES

-0.39

isierung

-0.38

icep

-0.38

some

-0.37

ology

-0.36

za

-0.36

ford

-0.36

cism

-0.35

RAL

-0.35

POSITIVE LOGITS

httphttps

0.70

IsContent

0.61

 OMITBAD

0.60

GTCX

0.60

IVEREF

0.59

 defaultstate

0.59

parsedMessage

0.58

jspb

0.57

oredCriteria

0.57

DockStyle

0.56

Activations Density 0.039%

references to specific parts or components in a technical or procedural text

conductive layer

No Comments

No Known Activations