INDEX

Explanations

references to geographical locations, specifically hills

oai_token-act-pair · gpt-3.5-turbo

references to hills and elevated terrains

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

uality

-1.01

ãĥ¯

-0.74

ually

-0.74

âĸĳâĸĳ

-0.67

 Consent

-0.66

~~~~

-0.65

 Attention

-0.65

ECA

-0.65

Mach

-0.65

 Role

-0.64

POSITIVE LOGITS

side

1.18

tops

0.99

 hill

0.94

top

0.91

 hills

0.89

frog

0.89

 slopes

0.87

stead

0.83

castle

0.82

bike

0.82

Activations Density 0.014%

references to geographical locations, specifically hills

references to hills and elevated terrains

No Comments

No Known Activations