INDEX

Explanations

phrases indicating a recommendation to watch something

oai_token-act-pair · gpt-3.5-turbo

references to watching videos or content

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

ctrl

-0.98

phi

-0.72

VEN

-0.69

ãĤ¨ãĥ«

-0.69

ãĥ´

-0.68

interstitial

-0.68

sembly

-0.67

ascal

-0.66

cffffcc

-0.66

 misunderstanding

-0.65

POSITIVE LOGITS

tower

1.26

 Watching

1.13

dog

1.04

dogs

0.98

 Watch

0.85

Watch

0.84

 Dogs

0.84

ing

0.82

watch

0.81

 WATCH

0.80

Activations Density 0.021%

phrases indicating a recommendation to watch something

references to watching videos or content

No Comments

No Known Activations