INDEX

Explanations

words related to discrimination, bias, and prejudice

oai_token-act-pair · gpt-3.5-turbo

terms associated with extreme or derogatory labeling and bias

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

aird

-0.78

stellar

-0.73

stable

-0.72

frames

-0.70

 Indigo

-0.69

quart

-0.68

Sync

-0.67

erald

-0.67

 Chrys

-0.66

oglobin

-0.65

POSITIVE LOGITS

 tactics

1.06

 intimidation

1.01

 blackmail

1.00

 extortion

0.90

 spying

0.89

 perpetrated

0.89

 retaliation

0.88

 accusations

0.88

 abuses

0.87

 threats

0.87

Activations Density 0.283%

words related to discrimination, bias, and prejudice

terms associated with extreme or derogatory labeling and bias

No Comments

No Known Activations