INDEX

Explanations

words related to something that is unfinished, incomplete, or has untapped potential

oai_token-act-pair · gpt-3.5-turbo

phrases that express a lack of trust or reliability

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

anwhile

-0.89

phrine

-0.76

SHIP

-0.75

 Pigs

-0.71

hyde

-0.70

cium

-0.70

å§«

-0.67

*/(

-0.67

byss

-0.67

)=(

-0.66

POSITIVE LOGITS

ruly

1.27

itled

1.27

ested

1.14

ribut

1.10

rans

1.09

race

1.09

oward

1.08

ravel

1.07

rave

1.07

enable

1.06

Activations Density 0.021%

words related to something that is unfinished, incomplete, or has untapped potential

phrases that express a lack of trust or reliability

No Comments

No Known Activations