INDEX

Explanations

objects or concepts related to cyber security, data privacy, or online anonymity

oai_token-act-pair · gpt-3.5-turbo

terms related to proxies and their applications or implications in various contexts

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

los

-0.76

 Luck

-0.75

 Folk

-0.71

 stressed

-0.69

heim

-0.69

laim

-0.67

enna

-0.66

fle

-0.66

 Bach

-0.66

 Till

-0.65

POSITIVE LOGITS

 proxy

4.15

 proxies

3.83

proxy

2.75

 Proxy

2.65

Proxy

2.14

 surrogate

1.32

 prox

1.25

 puppet

1.12

 conduit

1.06

 surrog

1.00

Activations Density 0.023%

objects or concepts related to cyber security, data privacy, or online anonymity

terms related to proxies and their applications or implications in various contexts

No Comments

No Known Activations