INDEX

Explanations

mentions of pressure or urging for action

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "pressed" and its variations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

ĪĴ

-0.72

cknowled

-0.67

flame

-0.64

 cliff

-0.63

 Lights

-0.63

BIT

-0.63

verty

-0.62

bows

-0.60

Â¯Â¯Â¯Â¯

-0.60

fin

-0.58

POSITIVE LOGITS

 pressed

1.12

pressed

0.93

press

0.90

 presses

0.87

entary

0.82

pressure

0.81

ioned

0.79

urized

0.77

roth

0.76

 pressing

0.75

Activations Density 0.012%

mentions of pressure or urging for action

occurrences of the word "pressed" and its variations

No Comments

No Known Activations