INDEX

Explanations

the presence of read and write commands in code snippets

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

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-res/layer_33/width_131k/average_l0_91

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.33.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

 дописавши

-0.53

RTEX

-0.53

ագրություններ

-0.52

BeginContext

-0.51

thâu

-0.49

 للمعارف

-0.48

 kaarangay

-0.47

ulario

-0.47

lapsingToolbar

-0.46

 redonda

-0.46

POSITIVE LOGITS

脚注の使い方

0.46

TestingModule

0.41

 classification

0.39

almaz

0.36

 filing

0.36

ByExample

0.36

 JSON

0.36

 puni

0.35

 المال

0.35

CreateModel

0.35

Activations Density 0.002%

the presence of read and write commands in code snippets

No Comments

No Known Activations