INDEX

Explanations

words and phrases that emphasize significance, uniqueness, complexity, and richness

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

descriptive qualities and concepts

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

ControllerAdvice

-0.42

either

-0.42

再來

-0.40

zelf

-0.39

someone

-0.38

 tendrás

-0.38

 either

-0.38

頗

-0.37

給你

-0.36

 itinéraires

-0.36

POSITIVE LOGITS

 surla

0.58

 Administrativna

0.56

 defaultstate

0.51

 Chwiliwch

0.50

arakat

0.47

"]));

0.47

InjectAttribute

0.46

ymce

0.45

цездатний

0.45

findpost

0.45

Activations Density 0.192%

words and phrases that emphasize significance, uniqueness, complexity, and richness

descriptive qualities and concepts

No Comments

No Known Activations