INDEX

Explanations

code-related constructs and data manipulations

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

 defStyleAttr

-0.76

EnableWeb

-0.66

 surla

-0.63

reportWebVitals

-0.62

SerializedName

-0.61

MessageOf

-0.60

TagMode

-0.60

LookAnd

-0.59

RegressionTest

-0.59

aarrggbb

-0.58

POSITIVE LOGITS

sub

0.36

PR

0.33

sub

0.33

«

0.33

stø

0.33

 dever

0.32

Pres

0.31

 Pres

0.31

Sub

0.31

udu

0.31

Activations Density 0.055%

code-related constructs and data manipulations

No Comments

No Known Activations