INDEX

Explanations

names of individuals involved in incidents or news articles

oai_token-act-pair · gpt-3.5-turbo

names of individuals involved in criminal incidents

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

olkien

-0.95

 optim

-0.85

soDeliveryDate

-0.83

pse

-0.82

 philosophers

-0.80

 UNESCO

-0.79

 bestselling

-0.79

ãĤ§

-0.78

 Kyoto

-0.77

 Philos

-0.77

POSITIVE LOGITS

esville

0.93

 Police

0.91

 police

0.87

endez

0.87

ville

0.87

 RCMP

0.87

 Pleasant

0.86

 COUNTY

0.85

ingham

0.84

Police

0.83

Activations Density 0.577%

names of individuals involved in incidents or news articles

names of individuals involved in criminal incidents

No Comments

No Known Activations