INDEX

Explanations

instances of 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 @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

ple

-0.81

edition

-0.78

pite

-0.77

warm

-0.72

fal

-0.72

cakes

-0.71

ectar

-0.68

ensical

-0.67

erves

-0.67

secondary

-0.66

POSITIVE LOGITS

 interviewed

1.23

 interviews

1.11

 interviewing

1.07

Interview

1.00

 interview

0.99

 Interview

0.95

 interviewer

0.80

VO

0.75

 Transcript

0.73

 polled

0.71

Activations Density 0.008%

instances of interviews

instances of the word "interviewed."

No Comments

No Known Activations