INDEX

Explanations

words related to comparisons or highlighting similarities between different entities

oai_token-act-pair · gpt-3.5-turbo

tokens representing the end of a document

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

Ire

-0.66

athan

-0.62

Redditor

-0.60

 COUR

-0.56

å£

-0.56

osc

-0.55

 mans

-0.54

ãĥ¼ãĥĨ

-0.54

ãģł

-0.54

å½

-0.54

POSITIVE LOGITS

pires

1.06

pire

1.03

cription

1.01

pired

1.00

ynchron

1.00

bestos

0.95

ync

0.95

 opposed

0.93

phy

0.92

semb

0.89

Activations Density 0.045%

words related to comparisons or highlighting similarities between different entities

tokens representing the end of a document

No Comments

No Known Activations