INDEX

Explanations

patterns of formatting and structure in textual data

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

quoted numbers and code

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

 hạ

-0.39

fficients

-0.35

 gesteld

-0.33

 gezet

-0.32

 Fakten

-0.29

 verschill

-0.29

 metropolitana

-0.28

 einzige

-0.28

 leggen

-0.28

 BytesLike

-0.28

POSITIVE LOGITS

uxxxx

0.91

 resourceCulture

0.68

enderror

0.66

 дописавши

0.66

LookAnd

0.65

 виправивши

0.60

 MainAxisSize

0.60

 autorytatywna

0.59

.*")]

0.57

setVerticalGroup

0.57

Activations Density 0.472%

patterns of formatting and structure in textual data

quoted numbers and code

No Comments

No Known Activations