INDEX

Explanations

phrases related to possession or belonging

oai_token-act-pair · gpt-3.5-turbo

references to a shared perspective or experience

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Negative Logits

puff

-0.71

bender

-0.70

_-

-0.68

CU

-0.66

 Goes

-0.64

Pac

-0.64

netflix

-0.63

 Ambro

-0.63

CENT

-0.62

 Schr

-0.62

POSITIVE LOGITS

selves

1.35

own

1.20

¥ŀ

0.96

 ourselves

0.92

 adversary

0.90

 adversaries

0.90

 respective

0.88

 fearless

0.87

 opponent

0.82

 nation

0.82

Activations Density 0.119%

phrases related to possession or belonging

references to a shared perspective or experience

No Known Activations