INDEX

Explanations

phrases related to legal matters or actions

oai_token-act-pair · gpt-3.5-turbo

instances of the word "def" in various contexts, often related to definitions or derogatory terms

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.1.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

externalActionCode

-0.69

 Archdemon

-0.68

WER

-0.65

åŃ

-0.63

ONES

-0.63

ppo

-0.62

 Galile

-0.61

 Franks

-0.60

sth

-0.60

DragonMagazine

-0.59

POSITIVE LOGITS

ibr

1.05

ensible

1.03

acements

0.95

acement

0.93

ector

0.92

inished

0.91

def

0.89

luent

0.87

def

0.86

initions

0.86

Activations Density 0.007%

phrases related to legal matters or actions

instances of the word "def" in various contexts, often related to definitions or derogatory terms

No Comments

No Known Activations