INDEX

Explanations

adjectives or nouns related to harm or damage

oai_token-act-pair · gpt-3.5-turbo

terms related to the concept of harm or harmfulness

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

quart

-0.79

Kings

-0.71

Hun

-0.70

eely

-0.68

ARCH

-0.66

Whe

-0.65

ebus

-0.65

peak

-0.65

TeX

-0.64

gran

-0.64

POSITIVE LOGITS

 harmful

1.15

 harm

1.03

 undermin

1.01

 harms

0.88

 endanger

0.85

 adolesc

0.85

 detrimental

0.85

 consequences

0.84

 contamin

0.80

 harming

0.80

Activations Density 0.008%

adjectives or nouns related to harm or damage

terms related to the concept of harm or harmfulness

No Comments

No Known Activations