INDEX

Explanations

the word "were" in various contexts

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

past state or action

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

itself

-0.60

ş

-0.58

 itself

-0.57

 Diaz

-0.57

 Solis

-0.54

اً

-0.53

is

-0.52

 Dietz

-0.51

HasIndex

-0.51

isk

-0.50

POSITIVE LOGITS

 were

0.77

 voltak

0.75

 mereka

0.74

 Were

0.72

were

0.71

 były

0.70

Were

0.68

 Mereka

0.68

Mereka

0.66

 byste

0.65

Activations Density 0.152%

the word "were" in various contexts

past state or action

No Comments

No Known Activations