INDEX

Explanations

words related to actions or attributes that are evaluative or judgmental

oai_token-act-pair · gpt-3.5-turbo

words that convey a sense of affirmation or positive acknowledgment

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

SE

-0.64

 walk

-0.63

BUS

-0.63

falls

-0.63

Bee

-0.61

 conf

-0.61

WS

-0.61

 grace

-0.60

 squid

-0.60

 flock

-0.59

POSITIVE LOGITS

ative

4.50

atives

3.13

atively

2.84

ATIVE

2.25

ativity

2.01

ational

1.69

atory

1.65

ation

1.49

atorial

1.48

ations

1.48

Activations Density 0.008%

words related to actions or attributes that are evaluative or judgmental

words that convey a sense of affirmation or positive acknowledgment

No Comments

No Known Activations