INDEX

Explanations

terms related to contradictory statements or differing situations

oai_token-act-pair · gpt-3.5-turbo

phrases that reference a situation or condition being described as typical or atypical

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

inders

-0.63

ĪĴ

-0.62

cair

-0.60

estern

-0.60

flush

-0.59

elin

-0.57

apons

-0.56

umbn

-0.55

Redd

-0.55

 colle

-0.55

POSITIVE LOGITS

 anymore

0.93

 everywhere

0.81

ional

0.77

pring

0.75

 whatsoever

0.74

AppData

0.74

for

0.70

ioned

0.70

 here

0.69

 elsewhere

0.69

Activations Density 0.104%

terms related to contradictory statements or differing situations

phrases that reference a situation or condition being described as typical or atypical

No Comments

No Known Activations