INDEX

Explanations

proper nouns or names of individuals, possibly related to news or publications

oai_token-act-pair · gpt-3.5-turbo

mentions of representatives or political figures

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

istically

-0.85

olean

-0.79

ahime

-0.76

phal

-0.73

glers

-0.71

 Brist

-0.71

 Cage

-0.64

 Finnish

-0.64

GGGGGGGG

-0.63

 Firefly

-0.62

POSITIVE LOGITS

rint

1.32

orters

1.20

orted

1.13

utation

1.11

orter

1.08

ository

1.08

utations

1.07

ositories

1.02

utable

0.96

ublic

0.94

Activations Density 0.012%

proper nouns or names of individuals, possibly related to news or publications

mentions of representatives or political figures

No Comments

No Known Activations