INDEX

Explanations

mentions of executives or leadership positions

oai_token-act-pair · gpt-3.5-turbo

references to executive roles or titles

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

 pinch

-0.76

 Hanson

-0.69

vier

-0.69

 bias

-0.67

 swell

-0.66

 Norn

-0.66

 negative

-0.66

 marked

-0.64

 swarm

-0.63

 Harm

-0.63

POSITIVE LOGITS

exec

4.34

Exec

1.96

 Exec

1.85

Executive

1.71

execute

1.59

administ

1.52

exe

1.24

 execut

1.21

 exec

1.19

CEO

1.14

Activations Density 0.027%

mentions of executives or leadership positions

references to executive roles or titles

No Comments

No Known Activations