INDEX

Explanations

instances of the word "message"

oai_token-act-pair · gpt-3.5-turbo

repeated references to messages in different contexts

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

engeance

-0.87

aughs

-0.80

arte

-0.72

ONSORED

-0.70

biased

-0.67

authorized

-0.66

itals

-0.66

orks

-0.65

ESCO

-0.65

auder

-0.64

POSITIVE LOGITS

 messages

1.11

Message

1.06

 message

0.97

message

0.97

 Messages

0.95

FontSize

0.89

 goodbye

0.87

boxes

0.83

box

0.82

msg

0.82

Activations Density 0.017%

instances of the word "message"

repeated references to messages in different contexts

No Comments

No Known Activations