INDEX

Explanations

references to physical locations, particularly hills

oai_token-act-pair · gpt-3.5-turbo

mentions of hills

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

uality

-0.88

ãĥ¯

-0.73

âĹ¼

-0.69

Mach

-0.68

âĸĳâĸĳ

-0.68

ECA

-0.67

~~~~

-0.67

ually

-0.67

 Attention

-0.66

 Marketable

-0.65

POSITIVE LOGITS

side

1.10

tops

1.05

 hill

1.00

frog

0.94

 hills

0.93

 slopes

0.92

top

0.91

stead

0.84

castle

0.79

hill

0.79

Activations Density 0.013%

references to physical locations, particularly hills

mentions of hills

No Comments

No Known Activations