INDEX

Explanations

words related to criminal activities, particularly arson

oai_token-act-pair · gpt-3.5-turbo

terms associated with arson and related criminal activities

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

avez

-0.75

 Presidents

-0.74

onse

-0.71

nir

-0.70

enance

-0.69

ux

-0.69

fortune

-0.67

advertisement

-0.63

POST

-0.63

ktop

-0.63

POSITIVE LOGITS

 arson

0.92

witch

0.84

emouth

0.83

©¶æ¥µ

0.75

 charcoal

0.75

 stove

0.74

 burner

0.73

cest

0.72

 bats

0.71

 rall

0.69

Activations Density 0.014%

words related to criminal activities, particularly arson

terms associated with arson and related criminal activities

No Comments

No Known Activations