INDEX

Explanations

phrases related to items being unpacked or unopened

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

SHIP

-0.77

ãĤ·ãĥ£

-0.68

 warr

-0.65

pieces

-0.64

piece

-0.64

 Lauder

-0.64

taboola

-0.64

MAX

-0.63

 dominion

-0.62

haus

-0.62

POSITIVE LOGITS

ractical

1.25

ublic

1.20

romising

1.18

unct

1.15

rompt

1.13

orter

1.11

resent

1.10

ossible

1.08

acked

1.06

ressed

1.04

Activations Density 0.024%

phrases related to items being unpacked or unopened

No Comments

No Known Activations