INDEX

Explanations

words related to abstract concepts or art

oai_token-act-pair · gpt-3.5-turbo

references to abstract concepts or representations

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

odder

-0.75

UNCH

-0.72

kins

-0.69

attering

-0.67

risome

-0.67

cture

-0.67

owship

-0.67

rimp

-0.67

ggle

-0.66

ICAN

-0.66

POSITIVE LOGITS

ions

1.21

matter

0.96

edly

0.93

edIn

0.86

stract

0.84

 Matter

0.84

ured

0.83

urally

0.82

 painter

0.80

 syntax

0.78

Activations Density 0.030%

words related to abstract concepts or art

references to abstract concepts or representations

No Comments

No Known Activations