INDEX

Explanations

years related to various events or activities

oai_token-act-pair · gpt-3.5-turbo

references to the years 1997 and 1998

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

hop

-0.77

giving

-0.75

ysc

-0.72

holder

-0.71

holders

-0.70

hatt

-0.69

tro

-0.69

edo

-0.66

groups

-0.66

houses

-0.65

POSITIVE LOGITS

ĸļ

0.80

ãĥĥãĥī

0.77

-'

0.77

ãĤ¦ãĤ¹

0.72

uncture

0.69

 onward

0.67

 Moines

0.64

ords

0.63

å¹

0.63

 Newsweek

0.62

Activations Density 0.024%

years related to various events or activities

references to the years 1997 and 1998

No Comments

No Known Activations