INDEX

Explanations

words related to negative emotions or situations, particularly fear and hatred

oai_token-act-pair · gpt-3.5-turbo

terms associated with fear or negative emotions

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 increments

-0.68

tom

-0.65

 attm

-0.64

ty

-0.62

 load

-0.62

trans

-0.59

album

-0.59

 Divide

-0.59

 composition

-0.58

Rew

-0.58

POSITIVE LOGITS

 feared

3.75

 dreaded

1.95

 fears

1.70

 hated

1.69

 fearing

1.68

 fear

1.64

 despised

1.60

 hoped

1.59

 disliked

1.53

 fearful

1.44

Activations Density 0.011%

words related to negative emotions or situations, particularly fear and hatred

terms associated with fear or negative emotions

No Comments

No Known Activations