INDEX

Explanations

phrases related to corruption and controversy

oai_token-act-pair · gpt-3.5-turbo

references to specific names and terms associated with certain entities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

 corrid

-0.95

 Forth

-0.74

 resil

-0.72

 Niet

-0.71

 lapt

-0.70

 behav

-0.69

 leaps

-0.69

 conduc

-0.67

uation

-0.66

falls

-0.65

POSITIVE LOGITS

ItemTracker

0.87

xual

0.84

netflix

0.83

venue

0.81

hammad

0.81

undle

0.79

tsky

0.77

regate

0.74

Sexual

0.74

isoft

0.74

Activations Density 0.024%

phrases related to corruption and controversy

references to specific names and terms associated with certain entities

No Comments

No Known Activations