INDEX

Explanations

phrases or words relating to explosive or impactful actions being executed

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate removal or disconnection

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

Export

-0.83

anners

-0.72

otive

-0.71

geist

-0.64

tone

-0.63

ilic

-0.62

elsius

-0.61

rite

-0.60

iosyncr

-0.60

ancial

-0.60

POSITIVE LOGITS

 Broadway

0.79

cffffcc

0.71

shore

0.64

shoot

0.63

season

0.63

ansas

0.62

dun

0.62

hand

0.61

ibaba

0.60

 waivers

0.60

Activations Density 0.036%

phrases or words relating to explosive or impactful actions being executed

phrases that indicate removal or disconnection

No Comments

No Known Activations