INDEX

Explanations

phrases related to investigating events or incidents

oai_token-act-pair · gpt-3.5-turbo

repeated references to a specific event or incident

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

iciency

-0.80

iculty

-0.70

keye

-0.68

gard

-0.67

ategories

-0.66

tan

-0.65

bitious

-0.65

usage

-0.63

solete

-0.63

uable

-0.63

POSITIVE LOGITS

 transpired

1.08

 unfolded

0.95

 unfolding

0.91

 unfold

0.90

 yesterday

0.89

 backstage

0.86

 aboard

0.83

 afterwards

0.79

 happened

0.79

 beforehand

0.78

Activations Density 0.076%

phrases related to investigating events or incidents

repeated references to a specific event or incident

No Comments

No Known Activations