INDEX

Explanations

programming constructs that indicate the end of loops or conditional statements

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

end of loops

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

FieldOffset

-0.39

DataPropertyName

-0.38

ض

-0.36

keyup

-0.34

Читати

-0.34

 frutta

-0.32

 clear

-0.32

aircraft

-0.31

Heb

-0.31

penup

-0.31

POSITIVE LOGITS

$__

2.98

($__

1.92

$__

1.62

 endwhile

1.20

endwhile

0.85

 autorytatywna

0.80

 }}">

0.73

 endforeach

0.72

 Dernière

0.70

SourceChecksum

0.70

Activations Density 0.000%

programming constructs that indicate the end of loops or conditional statements

end of loops

No Comments

No Known Activations