INDEX

Explanations

greetings and instructions for action

oai_token-act-pair · gpt-3.5-turbo

requests or commands for action

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

-0.76

arthed

-0.75

bent

-0.71

lings

-0.68

é¾

-0.67

byss

-0.67

Balt

-0.67

aler

-0.67

Aust

-0.66

Osw

-0.65

POSITIVE LOGITS

 forgive

1.22

 enable

1.17

 note

1.16

 refrain

1.14

 excuse

1.13

 pardon

1.07

 consider

1.06

 refer

1.05

don

1.04

 remember

1.00

Activations Density 0.046%

greetings and instructions for action

requests or commands for action

No Comments

No Known Activations