INDEX

Explanations

mentions of Twitter usernames

oai_token-act-pair · gpt-3.5-turbo

sequences of underscores or placeholders for usernames or tags

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

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

 screenings

-0.83

 reception

-0.81

aud

-0.77

 screening

-0.71

 chlorine

-0.71

 manned

-0.70

 divers

-0.69

 bounce

-0.69

ric

-0.68

 Manson

-0.68

POSITIVE LOGITS

ebook

1.32

chance

1.26

EStreamFrame

1.08

vs

1.08

dust

1.04

RAW

1.04

must

1.02

SOURCE

1.00

deck

1.00

Activations Density 0.020%

mentions of Twitter usernames

sequences of underscores or placeholders for usernames or tags

No Comments

No Known Activations