INDEX

Explanations

mentions of being impressed by something or someone

oai_token-act-pair · gpt-3.5-turbo

expressions of admiration or being impressed

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

yl

-0.62

race

-0.61

BD

-0.60

="#

-0.59

access

-0.58

most

-0.58

 Place

-0.57

 Thirty

-0.56

Ban

-0.56

 Theft

-0.56

POSITIVE LOGITS

 impressed

3.59

 amazed

1.98

 intrigued

1.92

 impress

1.81

 pleased

1.80

 amused

1.78

 surprised

1.70

 dazz

1.65

 fascinated

1.64

 astonished

1.61

Activations Density 0.007%

mentions of being impressed by something or someone

expressions of admiration or being impressed

No Comments

No Known Activations