INDEX

Explanations

percentages

oai_token-act-pair · gpt-3.5-turbo

references to percentages

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

 firsthand

-0.67

 revoked

-0.63

 violated

-0.62

 backstage

-0.62

 warr

-0.61

 rewrite

-0.61

 postseason

-0.61

ãģł

-0.58

HF

-0.58

 offseason

-0.58

POSITIVE LOGITS

cent

4.79

cence

2.48

Cent

2.16

CENT

2.03

center

1.80

 cent

1.73

 Cent

1.50

 CENT

1.30

cing

1.29

centric

1.28

Activations Density 0.005%

percentages

references to percentages

No Comments

No Known Activations