INDEX

Explanations

non-English characters, potentially indicating a focus on a specific language or encoding scheme

oai_token-act-pair · gpt-3.5-turbo

special characters and potentially malformed text elements

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

 intersections

-0.69

eways

-0.65

merce

-0.63

 closures

-0.62

 Targ

-0.59

horn

-0.58

xus

-0.58

 Riley

-0.56

voy

-0.56

 shapes

-0.56

POSITIVE LOGITS

ľ

1.22

´

1.11

Ħ

1.08

Ð¸

1.07

ł

1.07

Ķ

1.06

ĺ

1.03

à¤¾

1.00

¤

0.99

°

0.97

Activations Density 0.035%

non-English characters, potentially indicating a focus on a specific language or encoding scheme

special characters and potentially malformed text elements

No Comments

No Known Activations