INDEX

Explanations

number ranges expressed in the format "start - end"

oai_token-act-pair · gpt-3.5-turbo

patterns related to year-long durations or comparisons

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

anium

-0.71

esson

-0.71

 Bullets

-0.70

SourceFile

-0.70

 Shooter

-0.67

eele

-0.65

keley

-0.65

 Mechdragon

-0.65

ãĥĦ

-0.64

idas

-0.64

POSITIVE LOGITS

based

1.10

sized

1.03

time

0.97

years

0.97

level

0.95

long

0.90

appointed

0.89

day

0.88

to

0.88

release

0.88

Activations Density 0.099%

number ranges expressed in the format "start - end"

patterns related to year-long durations or comparisons

No Comments

No Known Activations