INDEX

Explanations

phrases related to providing or lacking information or evidence in a structured manner

oai_token-act-pair · gpt-3.5-turbo

references to judicial processes or legal testimony

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Lithuan

-0.73

utenberg

-0.56

 Jupiter

-0.56

 circled

-0.56

 sandwic

-0.54

 envelope

-0.54

 Auckland

-0.54

loo

-0.54

 shortly

-0.53

 Venus

-0.51

POSITIVE LOGITS

nor

1.71

 anymore

1.42

 slightest

1.25

nor

1.23

 whatsoever

1.18

 anything

1.14

any

1.12

soever

1.05

anything

1.04

ught

1.03

Activations Density 0.986%

phrases related to providing or lacking information or evidence in a structured manner

references to judicial processes or legal testimony

No Comments

No Known Activations