INDEX

Explanations

references to the prefix "ex-" indicating something former or existing in the past

oai_token-act-pair · gpt-3.5-turbo

references to individuals who are labeled as "ex" followed by a noun, indicating their previous roles or relationships

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 GOODMAN

-0.81

 Drawn

-0.69

ãĤ¤ãĥĪ

-0.67

OUP

-0.66

 Sabha

-0.65

arity

-0.64

 Fever

-0.63

 Flavoring

-0.62

 Sawyer

-0.60

 halls

-0.60

POSITIVE LOGITS

clamation

1.43

orbit

1.24

uber

1.21

ogenous

1.19

terior

1.10

tern

1.08

clud

1.07

clus

1.07

communication

1.06

portation

1.05

Activations Density 0.015%

references to the prefix "ex-" indicating something former or existing in the past

references to individuals who are labeled as "ex" followed by a noun, indicating their previous roles or relationships

No Comments

No Known Activations