INDEX

Explanations

phrases describing actions or behaviors of individuals in different situations

oai_token-act-pair · gpt-3.5-turbo

actions related to illegal or inappropriate behavior

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

endif

-0.79

etheless

-0.78

soever

-0.75

nown

-0.74

atari

-0.74

fortunately

-0.70

ciating

-0.69

Alternatively

-0.67

cking

-0.67

inis

-0.65

POSITIVE LOGITS

 sake

0.96

 purposes

0.87

 improper

0.83

 illegal

0.79

 reasons

0.78

 violations

0.73

 nonviolent

0.72

 unpopular

0.72

 unlawful

0.69

 improperly

0.65

Activations Density 0.165%

phrases describing actions or behaviors of individuals in different situations

actions related to illegal or inappropriate behavior

No Comments

No Known Activations