INDEX

Explanations

apologies or laments

oai_token-act-pair · gpt-3.5-turbo

instances of the word "sorry."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 helicop

-0.69

ccording

-0.68

rower

-0.67

krit

-0.67

lav

-0.66

 holistic

-0.65

tein

-0.65

minecraft

-0.64

vet

-0.64

ief

-0.64

POSITIVE LOGITS

 sorry

1.06

 Sorry

0.91

sorry

0.87

Sorry

0.85

 excuse

0.82

 pardon

0.79

GES

0.75

 Guys

0.72

 Ladies

0.70

 apologies

0.69

Activations Density 0.009%

apologies or laments

instances of the word "sorry."

No Comments

No Known Activations