INDEX

Explanations

news-related entities or references

oai_token-act-pair · gpt-3.5-turbo

mentions of the CNN news network

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

sole

-0.72

laus

-0.72

irts

-0.71

powers

-0.71

mot

-0.70

vol

-0.70

gging

-0.68

price

-0.67

minded

-0.67

 periods

-0.66

POSITIVE LOGITS

CNN

1.29

Money

0.87

ienne

0.86

assetsadobe

0.84

pedia

0.83

CNN

0.83

 Journalists

0.80

ILLE

0.79

icates

0.77

Leaks

0.75

Activations Density 0.006%

news-related entities or references

mentions of the CNN news network

No Comments

No Known Activations