INDEX

Explanations

structured data presentations, likely in table format

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

tables and structured text

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

ThroughAttribute

-0.64

 estekak

-0.62

 חיצוניים

-0.61

dflare

-0.60

 intStringLen

-0.60

ResumeLayout

-0.59

 otomatig

-0.58

migrationBuilder

-0.58

 ligiloj

-0.57

StructEnd

-0.56

POSITIVE LOGITS

newtheorem

0.39

Datuak

0.37

ähr

0.37

 Grand

0.33

<tr>

0.33

setVerticalGroup

0.33

<?

0.30

 Dearborn

0.30

 Fort

0.29

 Douglas

0.29

Activations Density 0.026%

structured data presentations, likely in table format

tables and structured text

No Comments

No Known Activations