INDEX

Explanations

mentions of a specific speaker or speech events

oai_token-act-pair · gpt-3.5-turbo

instances of the word "Spe" and variations related to speech or speaking

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

 dough

-0.76

 addictive

-0.74

 detachment

-0.72

 retro

-0.70

 bonus

-0.70

 motel

-0.69

 loop

-0.68

 franchises

-0.67

 addict

-0.67

 Jamaica

-0.66

POSITIVE LOGITS

Spe

4.03

spe

1.91

speech

1.83

Spe

1.61

 Speak

1.48

 Speech

1.46

Talk

1.41

speak

1.38

 Speaker

1.35

 speaker

1.28

Activations Density 0.012%

mentions of a specific speaker or speech events

instances of the word "Spe" and variations related to speech or speaking

No Comments

No Known Activations