INDEX

Explanations

phrases expressing opinions or beliefs

oai_token-act-pair · gpt-3.5-turbo

clauses or phrases that indicate belief, doubt, or opinion statements

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

ãĥĦ

-0.73

ĸļå£«

-0.65

flag

-0.64

ActionCode

-0.64

EStream

-0.64

Said

-0.64

efe

-0.64

ocative

-0.64

tained

-0.63

Redditor

-0.63

POSITIVE LOGITS

 nowadays

1.35

 nobody

1.14

 there

1.10

 they

1.09

 alot

1.06

 lately

1.04

 although

1.03

 everybody

1.02

 whenever

1.02

 whereas

1.02

Activations Density 0.447%

phrases expressing opinions or beliefs

clauses or phrases that indicate belief, doubt, or opinion statements

No Comments

No Known Activations