INDEX

Explanations

terms related to criminal activities and individuals involved in criminal acts

oai_token-act-pair · gpt-3.5-turbo

references to criminality and related terms

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

oration

-0.75

tz

-0.69

oute

-0.68

Ã©e

-0.66

ITE

-0.64

reci

-0.64

 stead

-0.62

BF

-0.62

 McMaster

-0.61

feed

-0.60

POSITIVE LOGITS

 criminals

3.58

 thieves

2.24

crim

2.19

 robbers

2.07

 murderers

2.05

 criminality

1.98

 offenders

1.97

 rapists

1.94

criminal

1.87

 perpetrators

1.87

Activations Density 0.025%

terms related to criminal activities and individuals involved in criminal acts

references to criminality and related terms

No Comments

No Known Activations