INDEX

Explanations

phrases related to infringement or violation

oai_token-act-pair · gpt-3.5-turbo

instances of the word "assess" or its variations indicating evaluation or judgment

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

 pept

-0.69

umbo

-0.65

express

-0.65

AE

-0.64

 recover

-0.62

doi

-0.61

Fal

-0.61

 anterior

-0.60

 migr

-0.60

POSITIVE LOGITS

assed

1.29

assing

1.18

asses

1.02

assy

0.87

 Sorce

0.85

byss

0.84

retty

0.81

daq

0.80

holes

0.79

itant

0.79

Activations Density 0.013%

phrases related to infringement or violation

instances of the word "assess" or its variations indicating evaluation or judgment

No Comments

No Known Activations