INDEX

Explanations

words related to deception or withholding information

oai_token-act-pair · gpt-3.5-turbo

words related to the concept of "dudes."

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

 sustained

-0.65

 stall

-0.65

 flap

-0.65

 hungry

-0.64

hill

-0.63

 chest

-0.63

PAT

-0.63

 scra

-0.62

 Atkinson

-0.61

 local

-0.61

POSITIVE LOGITS

udes

4.70

ude

3.24

uding

2.17

uded

1.94

usions

1.62

usion

1.39

uders

1.27

usive

1.20

uder

1.15

ud

1.05

Activations Density 0.010%

words related to deception or withholding information

words related to the concept of "dudes."

No Comments

No Known Activations