INDEX

Explanations

words related to negative consequences, such as death, suicide, blames, and disease

oai_token-act-pair · gpt-3.5-turbo

references to mortality and dangerous situations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

anooga

-0.72

 dashed

-0.71

 continued

-0.71

 ensued

-0.69

urther

-0.68

 awaited

-0.66

 overcame

-0.64

 explored

-0.64

 commenced

-0.63

 grasped

-0.63

POSITIVE LOGITS

 nowadays

1.04

 yearly

1.01

 everyday

0.99

 daily

0.97

 regularly

0.96

 annually

0.86

 nightly

0.84

advertisement

0.82

ometimes

0.82

 constantly

0.80

Activations Density 1.100%

words related to negative consequences, such as death, suicide, blames, and disease

references to mortality and dangerous situations

No Comments

No Known Activations