INDEX

Explanations

text fragments that include non-English characters and symbols like "Ð" and "Ñ", possibly related to a specific language or encoding

oai_token-act-pair · gpt-3.5-turbo

non-English characters or potentially corrupted text

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

aday

-0.76

ORED

-0.73

iday

-0.67

Bun

-0.66

nels

-0.65

abee

-0.65

ITNESS

-0.64

igan

-0.63

 Starr

-0.62

Walker

-0.62

POSITIVE LOGITS

Ð

1.55

Ð

1.25

¾

1.15

ĳ

1.15

ľ

1.12

ļ

1.12

Ķ

1.08

ķ

1.08

Ð½

1.07

ŀ

1.04

Activations Density 0.005%

text fragments that include non-English characters and symbols like "Ð" and "Ñ", possibly related to a specific language or encoding

non-English characters or potentially corrupted text

No Comments

No Known Activations