INDEX

Explanations

warnings or cautions expressed in texts

oai_token-act-pair · gpt-3.5-turbo

warnings or cautions about potential risks

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

cess

-0.87

fab

-0.73

arp

-0.72

rid

-0.71

ID

-0.69

Doctor

-0.68

mut

-0.67

func

-0.67

bernatorial

-0.66

ater

-0.66

POSITIVE LOGITS

 beware

1.12

 flock

0.94

 Beware

0.86

 lest

0.79

rums

0.78

eware

0.77

theless

0.74

 wary

0.73

 heed

0.71

ashtra

0.71

Activations Density 0.029%

warnings or cautions expressed in texts

warnings or cautions about potential risks

No Comments

No Known Activations