INDEX

Explanations

phrases related to opting out or agreeing in communications

oai_token-act-pair · gpt-3.5-turbo

phrases indicating user options and permissions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 Mahjong

-0.62

 Denis

-0.58

say

-0.57

 Shim

-0.57

 Sabb

-0.57

Apps

-0.55

Ore

-0.55

Dia

-0.54

Shell

-0.53

 Sandwich

-0.53

POSITIVE LOGITS

 someday

0.88

ivably

0.76

chev

0.76

ewski

0.75

TAIN

0.71

loo

0.70

 depending

0.69

rued

0.68

 tempted

0.67

depending

0.67

Activations Density 0.220%

phrases related to opting out or agreeing in communications

phrases indicating user options and permissions

No Comments

No Known Activations