INDEX

Explanations

references to interviews

oai_token-act-pair · gpt-3.5-turbo

instances of the word "interviewed"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

flow

-0.69

BIT

-0.69

complex

-0.66

 limitation

-0.65

few

-0.65

 initiation

-0.64

board

-0.64

birth

-0.62

nings

-0.62

ship

-0.62

POSITIVE LOGITS

OHN

0.88

 extensively

0.86

 interviewed

0.84

 polled

0.77

 withd

0.77

iscons

0.77

ÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤ

0.74

 Photographer

0.73

hap

0.71

 citiz

0.71

Activations Density 0.034%

references to interviews

instances of the word "interviewed"

No Comments

No Known Activations