INDEX

Explanations

phrases with legal or adversarial connotations

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 prosec

-0.81

 princ

-0.77

 citiz

-0.76

 commissions

-0.75

¥ŀ

-0.74

 skelet

-0.74

 censored

-0.73

 obser

-0.73

 lifes

-0.72

 newcom

-0.70

POSITIVE LOGITS

"[

2.00

"(

1.91

1.90

He

1.69

"'

1.66

"...

1.57

Asked

1.49

Instead

1.48

She

1.46

His

1.44

Activations Density 0.344%

phrases with legal or adversarial connotations

No Comments

No Known Activations