INDEX

Explanations

references to time coincidences or events happening simultaneously

oai_token-act-pair · gpt-3.5-turbo

references to specific days and times

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

efully

-0.68

emort

-0.68

seek

-0.67

ntil

-0.64

chieve

-0.64

better

-0.63

err

-0.63

erk

-0.62

kind

-0.61

itus

-0.61

POSITIVE LOGITS

 that

0.97

as

0.88

AS

0.73

 THAT

0.66

 Prohibition

0.63

 Kirin

0.63

 Ivanka

0.61

as

0.57

 Newsweek

0.56

that

0.56

Activations Density 0.037%

references to time coincidences or events happening simultaneously

references to specific days and times

No Comments

No Known Activations