INDEX

Explanations

phrases with the structure "top-something"

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate rankings or top-tier status

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

Dia

-0.67

 Sawyer

-0.62

 Strait

-0.61

Syl

-0.60

 spells

-0.60

 vigil

-0.60

 Parenthood

-0.59

 Manual

-0.59

 Illum

-0.59

ITED

-0.59

POSITIVE LOGITS

level

1.25

tier

1.22

ranking

1.18

ranked

1.11

selling

1.09

class

1.08

lining

1.07

performing

1.04

heavy

1.02

priority

1.02

Activations Density 0.030%

phrases with the structure "top-something"

phrases that indicate rankings or top-tier status

No Comments

No Known Activations