INDEX

Explanations

patterns or sequences that appear structured or formulaic, possibly related to mathematical or programming syntax

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

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-res/layer_20/width_131k/average_l0_114

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

 ModelExpression

-0.72

awakeFromNib

-0.66

 /\.(

-0.58

gameserver

-0.55

LookAnd

-0.54

aarrggbb

-0.54

windowFixed

-0.52

 SwitchCompat

-0.50

 HttpNotFound

-0.50

webElementXpaths

-0.50

POSITIVE LOGITS

 nocturno

0.38

 tuvimos

0.37

工程

0.36

 monitored

0.35

 isti

0.35

 dosta

0.35

 récord

0.35

 progress

0.34

 rendering

0.34

 niveles

0.34

Activations Density 0.002%

patterns or sequences that appear structured or formulaic, possibly related to mathematical or programming syntax

No Comments

No Known Activations