INDEX

Explanations

phrases indicating bias and unfairness

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

not always

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

ânsito

-0.31

ly

-0.29

ēj

-0.29

 disfrut

-0.29

almost

-0.28

ěji

-0.28

 Schie

-0.28

 Forschung

-0.28

ādi

-0.27

 réputation

-0.27

POSITIVE LOGITS

Personendaten

0.75

 invokingState

0.73

 चीज़ों

0.61

errHandler

0.58

StoryboardSegue

0.56

 pulito

0.54

󠁴

0.53

Transkript

0.53

 аккура

0.52

 lenker

0.52

Activations Density 0.188%

phrases indicating bias and unfairness

not always

No Comments

No Known Activations