INDEX

Explanations

references to individuals working as clerks

oai_token-act-pair · gpt-3.5-turbo

words related to clerical positions or roles

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

 Palestin

-0.81

 Azerb

-0.74

uld

-0.72

 Fighting

-0.70

risk

-0.69

 Ottawa

-0.68

çīĪ

-0.66

 Caucasus

-0.65

NAT

-0.64

raid

-0.63

POSITIVE LOGITS

 clerk

1.53

 clerks

1.34

 Clerk

1.28

wright

0.90

ÃŃa

0.83

rette

0.81

inson

0.79

iffs

0.76

esses

0.75

 drawer

0.74

Activations Density 0.005%

references to individuals working as clerks

words related to clerical positions or roles

No Comments

No Known Activations