INDEX

Explanations

racially-charged terms and expressions

oai_token-act-pair · gpt-3.5-turbo

terms related to racism and associated negative descriptors

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ITNESS

-0.93

tis

-0.78

pad

-0.77

mits

-0.76

earchers

-0.76

itness

-0.75

nown

-0.74

arios

-0.73

eenth

-0.73

icular

-0.72

POSITIVE LOGITS

 slurs

1.29

 bigot

1.13

tir

1.07

 slur

1.06

 prejudice

1.01

 racist

1.00

 stereotype

0.99

 stereotypes

0.96

 supremacist

0.96

 racists

0.94

Activations Density 0.068%

racially-charged terms and expressions

terms related to racism and associated negative descriptors

No Comments

No Known Activations