INDEX

Explanations

occurrences of numerical data and related placeholders in structured formats

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B @ 21-gemmascope-res-16k

Configuration

google/gemma-scope-9b-pt-res/layer_21/width_16k/average_l0_129

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.21.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

AndEndTag

-0.91

تقاوى

-0.76

GOTREF

-0.76

 pinulongan

-0.72

aarrggbb

-0.72

EndContext

-0.71

DockStyle

-0.70

Diweddarwch

-0.69

хьтан

-0.68

 ModelExpression

-0.68

POSITIVE LOGITS

 again

0.43

 Again

0.42

Again

0.41

 Another

0.35

 señ

0.35

 another

0.35

 novamente

0.35

 taas

0.35

 yine

0.34

Another

0.33

Activations Density 1.430%

occurrences of numerical data and related placeholders in structured formats

No Comments

No Known Activations