INDEX

Explanations

politically related terms, particularly focusing on party affiliations

oai_token-act-pair · gpt-3.5-turbo

references to political parties and gender

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

mun

-0.76

 Stard

-0.67

adobe

-0.63

ANN

-0.61

 RELE

-0.61

 WARN

-0.60

 Producer

-0.58

abc

-0.57

atana

-0.57

andestine

-0.57

POSITIVE LOGITS

 counterpart

0.88

 counterparts

0.85

 equivalents

0.76

itto

0.68

 versions

0.66

 versa

0.65

 flakes

0.65

Ð´

0.65

 captivity

0.64

ngth

0.64

Activations Density 0.325%

politically related terms, particularly focusing on party affiliations

references to political parties and gender

No Comments

No Known Activations