INDEX

Explanations

words or phrases with the prefix "unw-" which may indicate something negative or opposing

oai_token-act-pair · gpt-3.5-turbo

mentions of the concept of being unwarranted or unjustified

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

phrine

-0.95

hyde

-0.88

cium

-0.86

senal

-0.81

uyomi

-0.76

ãĤ¼

-0.76

anwhile

-0.75

 hemor

-0.72

ĺħ

-0.70

 Defenders

-0.70

POSITIVE LOGITS

arranted

1.03

ritten

1.03

ashed

0.98

inding

0.93

irth

0.91

itt

0.91

ield

0.91

avering

0.88

onder

0.87

ashington

0.87

Activations Density 0.006%

words or phrases with the prefix "unw-" which may indicate something negative or opposing

mentions of the concept of being unwarranted or unjustified

No Comments

No Known Activations