INDEX

Explanations

language indicating caution or needing to be wary

oai_token-act-pair · gpt-3.5-turbo

references to the concept of being cautious or careful

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

NZ

-0.80

CVE

-0.77

upon

-0.73

Wars

-0.71

flat

-0.69

Apple

-0.68

NF

-0.67

Haunted

-0.67

Phones

-0.67

SN

-0.66

POSITIVE LOGITS

tarian

0.87

 scrutiny

0.86

 calibr

0.85

tarians

0.79

 enough

0.78

taker

0.74

 careful

0.72

ness

0.70

 empir

0.67

 deliber

0.66

Activations Density 0.014%

language indicating caution or needing to be wary

references to the concept of being cautious or careful

No Comments

No Known Activations