INDEX

Explanations

phrases related to software updates or new features being added

oai_token-act-pair · gpt-3.5-turbo

common phrases and conjunctions that connect elements of a discussion or list

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

¬¼

-0.74

iculty

-0.74

exit

-0.71

ety

-0.71

acht

-0.68

eur

-0.68

oni

-0.67

ilege

-0.67

iet

-0.66

oler

-0.66

POSITIVE LOGITS

 namely

1.59

 including

1.34

 Including

1.10

viz

1.08

 ranging

1.07

including

1.02

excluding

0.98

 excluding

0.95

 includ

0.94

 notably

0.94

Activations Density 0.600%

phrases related to software updates or new features being added

common phrases and conjunctions that connect elements of a discussion or list

No Comments

No Known Activations