INDEX

Explanations

references to physical rooms

oai_token-act-pair · gpt-3.5-turbo

references to physical spaces described as "rooms."

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

Iss

-0.70

 reconnect

-0.68

thood

-0.65

 vengeance

-0.65

ename

-0.64

 rebellious

-0.63

 neurot

-0.62

usterity

-0.61

Uz

-0.61

haps

-0.61

POSITIVE LOGITS

room

1.26

 room

1.22

Room

1.10

 Room

1.07

 rooms

1.04

rooms

0.89

 Rooms

0.89

bell

0.88

bryce

0.82

 floor

0.81

Activations Density 0.020%

references to physical rooms

references to physical spaces described as "rooms."

No Comments

No Known Activations