INDEX

Explanations

superlatives or words that signify being the best or highest in a certain context

oai_token-act-pair · gpt-3.5-turbo

terms related to leadership and authority figures

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 micro

-0.68

 bitters

-0.67

tap

-0.65

 modified

-0.65

 scanning

-0.64

 port

-0.63

 Universal

-0.63

 mash

-0.63

 ether

-0.63

fin

-0.63

POSITIVE LOGITS

leading

3.78

leader

3.43

leaders

3.42

lead

2.59

Leader

1.36

responsible

1.18

represent

1.14

chief

1.11

 Leading

1.08

led

1.07

Activations Density 0.014%

superlatives or words that signify being the best or highest in a certain context

terms related to leadership and authority figures

No Comments

No Known Activations