INDEX

Explanations

words related to harm or danger

oai_token-act-pair · gpt-3.5-turbo

terms related to harm and negative impacts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

ãĤº

-0.80

 Pens

-0.78

Tools

-0.68

 excav

-0.67

Activity

-0.65

mine

-0.65

Screen

-0.65

Poc

-0.65

Ig

-0.64

Enlarge

-0.63

POSITIVE LOGITS

 choke

1.82

harm

1.57

 dizz

1.35

 staggered

1.34

 bankrupt

1.33

 cant

1.31

 choked

1.28

 choking

1.27

 stagger

1.24

 inco

1.19

Activations Density 0.047%

words related to harm or danger

terms related to harm and negative impacts

No Comments

No Known Activations