INDEX

Explanations

commands or suggestions prompting the reader to try something

oai_token-act-pair · gpt-3.5-turbo

phrases encouraging attempts or efforts to engage with something

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

goers

-0.74

rone

-0.73

rors

-0.71

arta

-0.68

ullah

-0.67

enary

-0.66

irable

-0.66

resent

-0.66

 concern

-0.64

eries

-0.64

POSITIVE LOGITS

 unsuccessfully

0.99

 experimenting

0.90

 contacting

0.83

out

0.82

 harder

0.82

outs

0.81

 imagining

0.78

 swapping

0.76

 messing

0.73

 putting

0.72

Activations Density 0.049%

commands or suggestions prompting the reader to try something

phrases encouraging attempts or efforts to engage with something

No Comments

No Known Activations