INDEX

Explanations

specific patterns indicating the beginning and end of news articles or reports

oai_token-act-pair · gpt-3.5-turbo

news reports and references to journalistic sources

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 endeavour

-0.93

 appre

-0.91

 fulfil

-0.81

 colours

-0.80

 colour

-0.78

 endeav

-0.76

 goto

-0.75

gements

-0.75

 advoc

-0.74

 clay

-0.73

POSITIVE LOGITS

Experts

1.29

Officials

1.28

According

1.25

Prosecutors

1.23

Authorities

1.21

Former

1.20

Sources

1.18

Investigators

1.17

WASHINGTON

1.15

Attorney

1.14

Activations Density 0.144%

specific patterns indicating the beginning and end of news articles or reports

news reports and references to journalistic sources

No Comments

No Known Activations