INDEX

Explanations

mentions of physical damage or harm in written text

oai_token-act-pair · gpt-3.5-turbo

references to damage or harm

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

zsche

-0.78

zee

-0.71

rams

-0.70

ramid

-0.66

atorial

-0.65

 Bars

-0.64

liner

-0.63

 Pitch

-0.62

 liner

-0.60

gent

-0.60

POSITIVE LOGITS

 inflicted

1.14

 damage

1.01

 mitigation

0.97

 wrought

0.95

damage

0.87

 havoc

0.81

 damaged

0.79

 incurred

0.78

 damages

0.76

 horm

0.75

Activations Density 0.020%

mentions of physical damage or harm in written text

references to damage or harm

No Comments

No Known Activations