INDEX

Explanations

references to television news channels

oai_token-act-pair · gpt-3.5-turbo

references to television networks or broadcasts

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

ussen

-0.73

0000000000000000

-0.73

hammad

-0.69

hend

-0.69

cific

-0.68

leans

-0.68

tha

-0.67

strument

-0.66

 otherwise

-0.66

notations

-0.65

POSITIVE LOGITS

ILLE

1.26

 Everywhere

0.88

trop

0.86

NZ

0.84

 Broadcasting

0.83

iolet

0.83

OS

0.80

NG

0.80

ISION

0.77

PG

0.74

Activations Density 0.021%

references to television news channels

references to television networks or broadcasts

No Comments

No Known Activations