INDEX

Explanations

phrases referring to specific entities or people

oai_token-act-pair · gpt-3.5-turbo

the word "the” and its repeated occurrences in the text

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

ornings

-0.88

doms

-0.82

 ãĤµãĥ¼ãĥĨãĤ£ãĥ¯ãĥ³

-0.78

rastructure

-0.77

ofi

-0.74

roleum

-0.73

|--

-0.73

apons

-0.72

tar

-0.72

İĭ

-0.72

POSITIVE LOGITS

 hashtag

1.54

 commenter

1.37

 tweet

1.23

 caption

1.23

 commenters

1.20

 subreddit

1.13

 meme

1.13

 Huffington

1.12

 prank

1.11

 offending

1.10

Activations Density 0.417%

phrases referring to specific entities or people

the word "the” and its repeated occurrences in the text

No Comments

No Known Activations