INDEX

Explanations

phrases related to successfully escaping or avoiding consequences

oai_token-act-pair · gpt-3.5-turbo

phrases that express the idea of evasion or escape

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

existence

-0.64

 Yamato

-0.63

 Merrill

-0.62

 Jinping

-0.61

onics

-0.60

Uz

-0.59

GB

-0.58

ZI

-0.57

IAL

-0.56

 Primary

-0.56

POSITIVE LOGITS

uced

0.74

esville

0.73

/+

0.68

 Riding

0.66

 Painting

0.65

 Pony

0.62

oaded

0.62

started

0.61

oglu

0.60

 legisl

0.59

Activations Density 0.026%

phrases related to successfully escaping or avoiding consequences

phrases that express the idea of evasion or escape

No Comments

No Known Activations