INDEX

Explanations

phrases related to negative or derogatory terms

oai_token-act-pair · gpt-3.5-turbo

terms associated with critique and negative characterization

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

ahon

-0.79

chwitz

-0.72

earch

-0.71

ensive

-0.71

large

-0.71

arbon

-0.70

ĳ

-0.70

range

-0.68

angan

-0.67

ascript

-0.67

POSITIVE LOGITS

 extraord

1.47

gery

1.05

esses

1.03

hood

1.00

ry

0.93

 archetype

0.91

liness

0.87

 persona

0.87

who

0.86

doms

0.85

Activations Density 0.360%

phrases related to negative or derogatory terms

terms associated with critique and negative characterization

No Comments

No Known Activations