INDEX

Explanations

years, specifically the year 2008

oai_token-act-pair · gpt-3.5-turbo

references to the year 2008

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

onite

-0.71

hma

-0.68

 ancest

-0.67

 unin

-0.66

unal

-0.65

Magikarp

-0.65

orate

-0.65

friends

-0.65

 rails

-0.64

lying

-0.64

POSITIVE LOGITS

å¹

0.97

-'

0.81

é¾

0.80

0.74

0.73

aeda

0.72

0.71

worthiness

0.67

0.65

ilton

0.65

Activations Density 0.017%

years, specifically the year 2008

references to the year 2008

No Comments

No Known Activations