INDEX

Explanations

references to programming and web development tasks

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

 anuales

-0.36

 annuelle

-0.34

 meille

-0.33

 disfrute

-0.33

 caballos

-0.33

 Abenteuer

-0.32

 gillar

-0.31

vuo

-0.30

 kanske

-0.30

 vuoden

-0.30

POSITIVE LOGITS

ſelf

0.87

ſicht

0.85

 faſt

0.84

iſchen

0.84

 zwiſchen

0.83

ſelves

0.83

niſſe

0.82

 ſche

0.81

iſten

0.81

 ſehen

0.80

Activations Density 0.012%

references to programming and web development tasks

No Comments

No Known Activations