INDEX

Explanations

news network names

oai_token-act-pair · gpt-3.5-turbo

references to news organizations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

dies

-0.71

tee

-0.71

 screws

-0.67

rocal

-0.65

 pend

-0.63

 Malfoy

-0.63

come

-0.63

abiding

-0.62

olester

-0.61

 appointing

-0.61

POSITIVE LOGITS

News

1.10

 News

0.98

news

0.86

NEWS

0.81

 Insight

0.81

 Wire

0.80

 correspondent

0.80

Politics

0.79

ople

0.78

 Networks

0.77

Activations Density 0.114%

news network names

references to news organizations

No Comments

No Known Activations