INDEX

Explanations

references to prohibition or restrictions

oai_token-act-pair · gpt-3.5-turbo

phrases that include the word "no."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

ortium

-0.69

iership

-0.68

minster

-0.65

 Catalog

-0.63

thora

-0.62

ulative

-0.61

uala

-0.60

istani

-0.59

ousand

-0.57

imir

-0.57

POSITIVE LOGITS

xious

1.38

va

1.13

vell

1.07

obs

1.05

 matter

1.01

ct

0.98

otrop

0.98

oses

0.97

vae

0.97

onday

0.95

Activations Density 0.049%

references to prohibition or restrictions

phrases that include the word "no."

No Comments

No Known Activations