INDEX

Explanations

Twitter usernames

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 striking

-0.78

 fulfilling

-0.77

 pressing

-0.75

 varied

-0.74

 shortage

-0.73

 sandwiches

-0.70

 alone

-0.67

 pacing

-0.67

.–

-0.67

 certificates

-0.67

POSITIVE LOGITS

Writ

1.35

1.31

Reports

1.14

Official

1.14

GOP

1.13

Coach

1.11

Real

1.10

CBC

1.10

Twe

1.08

Ide

1.08

Activations Density 0.094%

Twitter usernames

No Comments

No Known Activations