INDEX

Explanations

percentages expressed as numerical values

oai_token-act-pair · gpt-3.5-turbo

numerical statistics related to percentage changes

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

bies

-0.67

pton

-0.64

©¶æ

-0.64

\\\\\\\\\\\\\\\\

-0.60

gotten

-0.58

 larg

-0.57

 prizes

-0.57

 neighb

-0.57

åİ

-0.57

ãĥ¤

-0.57

POSITIVE LOGITS

xual

0.96

bps

0.80

asin

0.74

+.

0.73

+,

0.73

ABV

0.72

fal

0.71

elsius

0.71

rowth

0.71

uary

0.70

Activations Density 0.085%

percentages expressed as numerical values

numerical statistics related to percentage changes

No Comments

No Known Activations