INDEX

Explanations

words related to various news topics or announcements

oai_token-act-pair · gpt-3.5-turbo

repeated references to news or news-related content

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

 randomized

-0.70

 calibration

-0.67

 marked

-0.67

 mapping

-0.66

 magnetic

-0.66

 quart

-0.65

-0.64

 determining

-0.64

 psychologist

-0.64

 compensated

-0.63

POSITIVE LOGITS

ews

5.22

ew

1.72

EW

1.22

ools

1.09

ewski

1.02

ovie

0.95

aughs

0.93

edia

0.88

ufact

0.88

ornings

0.86

Activations Density 0.010%

words related to various news topics or announcements

repeated references to news or news-related content

No Comments

No Known Activations