INDEX

Explanations

phrases indicating hope, enjoyment, or support from the reader

oai_token-act-pair · gpt-3.5-turbo

expressions of encouragement and positive sentiment towards an audience

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ãĤ´ãĥ³

-0.74

£ı

-0.72

ĨĴ

-0.66

 Canaver

-0.64

 purportedly

-0.60

itta

-0.59

mere

-0.59

itol

-0.58

Sold

-0.58

ELD

-0.57

POSITIVE LOGITS

 enjoyed

1.24

 enjoy

1.18

'll

1.05

 guys

1.03

 liked

1.00

 Enjoy

1.00

 appreciate

0.94

're

0.90

 appreciated

0.84

 enjoys

0.82

Activations Density 0.086%

phrases indicating hope, enjoyment, or support from the reader

expressions of encouragement and positive sentiment towards an audience

No Comments

No Known Activations