INDEX

Explanations

the year "2008"

oai_token-act-pair · gpt-3.5-turbo

mentions of 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 @ 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

 ancest

-0.77

ymes

-0.67

heid

-0.64

Magikarp

-0.64

venge

-0.63

detail

-0.63

ancy

-0.61

 unpre

-0.60

 Rooney

-0.60

afort

-0.60

POSITIVE LOGITS

å¹

0.75

-'

0.69

finals

0.68

yrinth

0.68

é¾

0.68

irection

0.67

wark

0.66

 Olympics

0.65

iyah

0.64

clinton

0.63

Activations Density 0.026%

the year "2008"

mentions of the year 2008

No Comments

No Known Activations