INDEX

Explanations

phrases related to physical acts of aggression or impact

oai_token-act-pair · gpt-3.5-turbo

references to the word "punch" 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 @ 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

uve

-0.92

abeth

-0.76

aird

-0.68

icter

-0.65

Private

-0.64

 Citizen

-0.64

 Neural

-0.64

udic

-0.62

rians

-0.61

 Archdemon

-0.61

POSITIVE LOGITS

bowl

1.25

 punches

0.87

 punch

0.84

aneers

0.80

bag

0.76

istani

0.75

outs

0.75

 punching

0.74

sticks

0.74

cart

0.74

Activations Density 0.014%

phrases related to physical acts of aggression or impact

references to the word "punch" in various contexts

No Comments

No Known Activations