INDEX

Explanations

phrases indicating emphasis or significance

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate discomfort or dissatisfaction in societal contexts

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

 Windsor

-0.72

clusive

-0.71

 Slash

-0.70

 Meaning

-0.63

 STOR

-0.63

 Koen

-0.60

 Heist

-0.59

 Closing

-0.59

 Pieces

-0.59

 pandemonium

-0.58

POSITIVE LOGITS

were

0.96

have

0.95

 reacted

0.93

had

0.88

should

0.85

are

0.85

 were

0.84

 cannot

0.84

who

0.83

 behaved

0.82

Activations Density 0.278%

phrases indicating emphasis or significance

phrases that indicate discomfort or dissatisfaction in societal contexts

No Comments

No Known Activations