INDEX

Explanations

mentions of specific names related to a controversial meeting or collusion

oai_token-act-pair · gpt-3.5-turbo

references to meetings and interactions with individuals connected to Russia

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

 Veter

-0.75

Pyr

-0.74

 delim

-0.73

 relegation

-0.71

WCS

-0.67

 Jagu

-0.67

ahime

-0.66

 Cork

-0.65

 Cary

-0.63

 à¨

-0.63

POSITIVE LOGITS

arnaev

1.00

 spying

0.86

ossier

0.86

FBI

0.84

 Emails

0.83

 dossier

0.83

ervatives

0.82

 undercover

0.81

 lied

0.80

 informants

0.79

Activations Density 0.152%

mentions of specific names related to a controversial meeting or collusion

references to meetings and interactions with individuals connected to Russia

No Comments

No Known Activations