INDEX

Explanations

references to time periods or durations, particularly emphasizing both past length and recentness

oai_token-act-pair · gpt-3.5-turbo

references to durations of time, particularly "years."

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

ople

-0.92

ess

-0.69

iquid

-0.67

anus

-0.64

apa

-0.64

hod

-0.64

 contag

-0.63

icum

-0.63

achus

-0.62

ussed

-0.61

POSITIVE LOGITS

ago

1.31

Ago

1.06

 olds

0.93

old

0.81

 hindsight

0.77

 Anniversary

0.75

 anniversary

0.74

 apiece

0.73

 overdue

0.71

 Warranty

0.71

Activations Density 0.087%

references to time periods or durations, particularly emphasizing both past length and recentness

references to durations of time, particularly "years."

No Comments

No Known Activations