INDEX

Explanations

words related to negative actions or behaviors such as "reprehensible" or "ugly."

oai_token-act-pair · gpt-3.5-turbo

terms associated with moral condemnation and criticism

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

glers

-0.89

ERY

-0.79

 Cage

-0.78

 Abyss

-0.78

ppo

-0.78

 Bruins

-0.77

 Ducks

-0.77

ggle

-0.73

ahime

-0.71

 Leap

-0.69

POSITIVE LOGITS

utations

1.39

rint

1.10

onse

1.08

ainted

1.07

uted

1.05

arations

1.03

rieve

1.02

lying

1.02

utation

1.02

ulsive

1.01

Activations Density 0.008%

words related to negative actions or behaviors such as "reprehensible" or "ugly."

terms associated with moral condemnation and criticism

No Comments

No Known Activations