INDEX

Explanations

words related to being inactive or unused

oai_token-act-pair · gpt-3.5-turbo

repeated mentions of inactivity or being idle

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 Ukrain

-0.81

obal

-0.74

obar

-0.71

imet

-0.71

gur

-0.70

 Bullets

-0.70

aled

-0.69

henko

-0.69

 iodine

-0.67

acea

-0.67

POSITIVE LOGITS

SPA

0.79

 Idle

0.78

ynes

0.76

steen

0.75

 chatter

0.73

pta

0.73

jad

0.72

borgh

0.72

 idle

0.71

stock

0.70

Activations Density 0.032%

words related to being inactive or unused

repeated mentions of inactivity or being idle

No Comments

No Known Activations