INDEX

Explanations

words related to reputation or representation

oai_token-act-pair · gpt-3.5-turbo

words or terms related to representation and reporters

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

istically

-0.73

scape

-0.66

gha

-0.62

gaard

-0.61

 Archdemon

-0.60

 EntityItem

-0.59

 Brist

-0.59

wings

-0.58

 Conditions

-0.57

=-=-=-=-

-0.57

POSITIVE LOGITS

ublic

1.52

resents

1.18

atri

1.04

orters

0.99

rint

0.96

ulse

0.95

ession

0.94

itude

0.87

rising

0.87

aint

0.86

Activations Density 0.019%

words related to reputation or representation

words or terms related to representation and reporters

No Comments

No Known Activations