INDEX

Explanations

terms related to identification and distinction

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

evident and detectable

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

flüs

-0.37

stufen

-0.35

endosi

-0.34

 leash

-0.33

 Nusa

-0.33

 cuerda

-0.32

ikations

-0.32

 Ghana

-0.32

 Tiempo

-0.32

 cordes

-0.32

POSITIVE LOGITS

OGND

0.63

 nahilalakip

0.56

+#+

0.54

 unmistakable

0.53



0.52

queryInterface

0.52

 unmistak

0.51

 iNdEx

0.50

cerpt

0.50

 transfieras

0.50

Activations Density 0.038%

terms related to identification and distinction

evident and detectable

No Comments

No Known Activations