INDEX

Explanations

words related to inciting anger or conflict

oai_token-act-pair · gpt-3.5-turbo

references to actions or events that incite conflict or tension

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

 Accounting

-0.88

ummer

-0.74

haul

-0.72

aum

-0.72

rake

-0.70

amacare

-0.70

ultz

-0.66

ewater

-0.66

oard

-0.66

fficiency

-0.65

POSITIVE LOGITS

 provocation

1.16

 provoke

1.02

 provoking

0.97

 bystanders

0.88

 provocative

0.88

 provoked

0.83

 prov

0.81

 aggression

0.79

 sidx

0.79

 escalation

0.79

Activations Density 0.033%

words related to inciting anger or conflict

references to actions or events that incite conflict or tension

No Comments

No Known Activations