INDEX

Explanations

"crummy" or negatively connotated words

oai_token-act-pair · gpt-3.5-turbo

references to the term "dummy" and related variants

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

ren

-0.69

Rio

-0.68

 Rogers

-0.67

 Forth

-0.66

 Hawks

-0.66

 Berk

-0.66

Rog

-0.65

 Atlanta

-0.64

 cove

-0.64

 proclamation

-0.64

POSITIVE LOGITS

ummy

4.23

ummies

2.61

umm

1.54

 mummy

1.53

ums

1.36

 dummy

1.23

ammed

1.20

ammy

1.19

umbo

1.12

umpy

1.08

Activations Density 0.020%

"crummy" or negatively connotated words

references to the term "dummy" and related variants

No Comments

No Known Activations