INDEX

Explanations

elements related to structured data or programming syntax

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

WIDTH of layout elements

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

 zijne

-0.60

aarrggbb

-0.59

 فريبيس

-0.57

 myſelf

-0.56

Tembelea

-0.55

Rüyada

-0.54

 onderdeel

-0.52

 للمعارف

-0.51

StructEnd

-0.51



-0.51

POSITIVE LOGITS

 again

0.64

again

0.52

Again

0.48

 Again

0.45

 опять

0.42

 yine

0.41

 taip

0.37

 nuevamente

0.36

 AGAIN

0.35

 novamente

0.35

Activations Density 0.004%

elements related to structured data or programming syntax

WIDTH of layout elements

No Comments

No Known Activations