INDEX

Explanations

phrases related to criminal activities, particularly crimes against humanity

oai_token-act-pair · gpt-3.5-turbo

phrases related to crimes committed against individuals or groups

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

soDeliveryDate

-1.10

natureconservancy

-0.90

chell

-0.87

çīĪ

-0.86

éĹĺ

-0.85

ERN

-0.84

 teasp

-0.83

hower

-0.83

OTUS

-0.83

}}}

-0.82

POSITIVE LOGITS

 civilians

1.02

 innocent

0.93

him

0.86

 humanity

0.85

 women

0.80

 unborn

0.79

 Humanity

0.79

 whom

0.76

 them

0.75

 anybody

0.73

Activations Density 0.049%

phrases related to criminal activities, particularly crimes against humanity

phrases related to crimes committed against individuals or groups

No Comments

No Known Activations