INDEX

Explanations

references to physical tables or surfaces

oai_token-act-pair · gpt-3.5-turbo

references to the word "table"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

vernment

-0.90

berto

-0.72

trak

-0.71

 Euras

-0.68

 Directorate

-0.65

ancock

-0.64

obbies

-0.64

esp

-0.63

irtual

-0.62

 Enhancement

-0.61

POSITIVE LOGITS

cloth

1.49

 table

1.03

tops

1.00

 tables

0.97

table

0.96

sheet

0.82

 TABLE

0.78

top

0.78

 scraps

0.76

stairs

0.75

Activations Density 0.012%

references to physical tables or surfaces

references to the word "table"

No Comments

No Known Activations