INDEX

Explanations

copyright and attribution information in texts

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

code comments and identifiers

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

 plus

-0.32

asInstanceOf

-0.30

 acting

-0.29

Polecam

-0.28

 utom

-0.28

<bos>

-0.27

 positive

-0.27

 anything

-0.27

 outside

-0.26

ürlich

-0.26

POSITIVE LOGITS



0.91

httphttps

0.87

Personensuche

0.85

 EconPapers

0.74

expandindo

0.71

DoubleQuotes

0.66

WebVitals

0.65

 defaultstate

0.65

BufferException

0.65

MigrationBuilder

0.65

Activations Density 0.002%

copyright and attribution information in texts

code comments and identifiers

No Comments

No Known Activations