INDEX

Explanations

instances of the word "Created" and variations of the term "Sent" in relation to document metadata

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

<eos>

-0.43

 still

-0.43

 mode

-0.41

<bos>

-0.41

 matching

-0.41

 back

-0.40

 others

-0.39

 extra

-0.39

 grip

-0.37

 revenge

-0.37

POSITIVE LOGITS

Created

1.05

Datuak

0.83

 للاسماء

0.80

 AssemblyCompany

0.79

SequentialGroup

0.79

 Created

0.77

MemoryWarning

0.77

 initComponents

0.76

+#+

0.75

 surla

0.75

Activations Density 0.000%

instances of the word "Created" and variations of the term "Sent" in relation to document metadata

No Comments

No Known Activations