INDEX

Explanations

phrases related to earning a spot or place in a competitive context

oai_token-act-pair · gpt-3.5-turbo

references to positions or rankings

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

rays

-0.64

earchers

-0.62

 weeds

-0.61

flix

-0.61

 hoops

-0.61

umb

-0.60

 rays

-0.59

 Turtles

-0.59

onduct

-0.58

 fallout

-0.58

POSITIVE LOGITS

 atop

1.01

holders

0.91

Reviewer

0.85

holder

0.84

bos

0.81

 among

0.80

buster

0.79

TRY

0.78

 slot

0.77

manship

0.76

Activations Density 0.091%

phrases related to earning a spot or place in a competitive context

references to positions or rankings

No Comments

No Known Activations