INDEX

Explanations

words related to gender and sexual orientation, as well as terms related to discrimination based on gender, sexual orientation, and race

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

Data Type

torch.float32

Hook Point

blocks.8.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

enium

-0.72

sonian

-0.72

bara

-0.71

rican

-0.71

kus

-0.71

 Wake

-0.69

bury

-0.68

hap

-0.67

 Springfield

-0.67

urion

-0.66

POSITIVE LOGITS

 discrimination

0.96

bending

0.93

 differences

0.91

 specificity

0.91

 affinity

0.90

 diversity

0.88

 slurs

0.86

 coded

0.85

blind

0.85

 affiliation

0.85

Activations Density 0.858%

words related to gender and sexual orientation, as well as terms related to discrimination based on gender, sexual orientation, and race

No Comments

No Known Activations