INDEX

Explanations

references to information leaks or breaches

oai_token-act-pair · gpt-3.5-turbo

references to information breaches or disclosures

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

ryu

-0.70

 Schw

-0.65

 Krish

-0.64

Sax

-0.64

eka

-0.63

Num

-0.63

 Cynthia

-0.62

Sue

-0.62

Sax

-0.62

song

-0.61

POSITIVE LOGITS

 leaks

3.96

 leaking

2.26

Leaks

2.24

 leaked

2.18

 leak

2.15

 leakage

1.84

 spills

1.75

 disclosures

1.48

 releases

1.44

 whistleblowers

1.41

Activations Density 0.011%

references to information leaks or breaches

references to information breaches or disclosures

No Comments

No Known Activations