INDEX

Explanations

phrases related to specific instructions or requests addressed to the reader

oai_token-act-pair · gpt-3.5-turbo

phrases that address or direct the reader

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

Georg

-0.67

 Sabha

-0.64

itol

-0.61

wikipedia

-0.60

 Louie

-0.60

bats

-0.58

©¶æ¥µ

-0.58

 Gamb

-0.58

Deg

-0.57

 Ambro

-0.57

POSITIVE LOGITS

're

1.42

 want

1.15

've

1.14

 decide

1.04

 encount

1.02

 wish

1.02

 wanna

1.01

 prefer

1.00

 intend

0.99

 haven

0.99

Activations Density 0.110%

phrases related to specific instructions or requests addressed to the reader

phrases that address or direct the reader

No Comments

No Known Activations