INDEX

Explanations

car classifications in racing contexts followed by mentions of a win

oai_token-act-pair · gpt-3.5-turbo

references to the "GT" entity, likely denoting a specific item or concept

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 venom

-0.74

fe

-0.68

 polio

-0.67

 Palace

-0.66

Ble

-0.64

Boh

-0.62

 Sanskrit

-0.62

 crab

-0.61

 Burns

-0.60

 contagious

-0.60

POSITIVE LOGITS

GT

3.54

gt

3.54

GG

1.73

GC

1.67

GV

1.46

GS

1.29

GN

1.23

GE

1.20

Gs

1.20

GP

1.16

Activations Density 0.024%

car classifications in racing contexts followed by mentions of a win

references to the "GT" entity, likely denoting a specific item or concept

No Comments

No Known Activations