INDEX

Explanations

punctuation marks and format indicators

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

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 esternos

-0.84

Cyfeiriadau

-0.76

ConstraintMaker

-0.74

 ſind

-0.65

 estekak

-0.64

CppMethod

-0.64

bootstrapcdn

-0.63

 שוליים

-0.63

клопе

-0.63

 oprot

-0.62

POSITIVE LOGITS

Personendaten

0.29

Everybody

0.29

<eos>

0.27

Ben

0.27

 Everybody

0.26

 mencionados

0.26

<blockquote>

0.26

Mysterious

0.25

↵↵

0.25

 benzina

0.24

Activations Density 0.019%

punctuation marks and format indicators

No Comments

No Known Activations