INDEX

Explanations

phrases indicating a recommendation to view content

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

news headlines that are labeled as "MUST WATCH."

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

acia

-0.71

ised

-0.64

 bleed

-0.63

bun

-0.62

nib

-0.61

 clay

-0.60

 halluc

-0.60

 envy

-0.58

 recl

-0.58

 amalg

-0.58

POSITIVE LOGITS

 WATCH

0.75

 VIDEOS

0.73

 Watching

0.69

esome

0.68

 IMAGES

0.68

------------------------------------------------

0.68

dog

0.67

...]

0.63

ARDS

0.62

degree

0.62

Activations Density 0.010%

phrases indicating a recommendation to view content

news headlines that are labeled as "MUST WATCH."

No Comments

No Known Activations