INDEX

Explanations

forms of the word "meditate" or closely related concepts

oai_token-act-pair · gpt-3.5-turbo

terms related to premeditation, particularly in violent contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

 fundament

-0.71

chnology

-0.69

IRE

-0.68

Â·Â·

-0.64

izons

-0.64

holders

-0.63

issues

-0.63

ordinate

-0.58

Filename

-0.58

Shares

-0.57

POSITIVE LOGITS

ieval

1.22

ication

1.18

ications

1.07

icated

1.01

iterranean

1.01

icative

0.99

icate

0.96

rill

0.90

achusetts

0.88

itated

0.86

Activations Density 0.009%

forms of the word "meditate" or closely related concepts

terms related to premeditation, particularly in violent contexts

No Comments

No Known Activations