INDEX

Explanations

references to the brand "Fire" in various contexts

oai_token-act-pair · gpt-3.5-turbo

references to the brand "Fire" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

ĸļ

-0.87

xus

-0.80

eter

-0.70

 Morg

-0.69

eday

-0.68

edy

-0.67

 Lans

-0.64

atem

-0.62

oulos

-0.62

eston

-0.60

POSITIVE LOGITS

flies

1.41

fighter

1.21

fighters

1.16

storm

1.15

fly

1.09

proof

1.03

fight

1.02

balls

1.02

birds

1.00

bird

1.00

Activations Density 0.035%

references to the brand "Fire" in various contexts

references to the brand "Fire" in various contexts

No Comments

No Known Activations