INDEX

Explanations

names associated with news articles or reports

oai_token-act-pair · gpt-3.5-turbo

hyphenated names or phrases

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

 operating

-0.71

 susp

-0.71

 aggregate

-0.68

 elim

-0.67

 NEXT

-0.67

 incub

-0.67

occ

-0.67

 hiber

-0.66

 procedural

-0.66

PRO

-0.66

POSITIVE LOGITS

chan

1.19

sama

1.17

style

1.14

kun

1.12

san

1.06

Jones

1.02

worthy

1.01

directed

0.99

induced

0.98

Young

0.96

Activations Density 0.034%

names associated with news articles or reports

hyphenated names or phrases

No Comments

No Known Activations