INDEX

Explanations

news clips or videos to watch

oai_token-act-pair · gpt-3.5-turbo

instances of the word "Watch" related to viewing content

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

bably

-0.76

 suscept

-0.68

emia

-0.68

uria

-0.67

 comprom

-0.65

pmwiki

-0.61

 inval

-0.61

recy

-0.61

ayers

-0.60

ctrl

-0.60

POSITIVE LOGITS

tower

1.16

dog

1.07

dogs

0.96

 Video

0.88

 Below

0.87

 clips

0.87

 footage

0.86

 Dogs

0.84

 Trailer

0.84

 below

0.83

Activations Density 0.035%

news clips or videos to watch

instances of the word "Watch" related to viewing content

No Comments

No Known Activations