INDEX

Explanations

words related to news or information

oai_token-act-pair · gpt-3.5-turbo

instances of the letter "n"

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

å§«

-0.69

gerald

-0.69

 compe

-0.64

rero

-0.63

 isEnabled

-0.62

awaru

-0.60

MODE

-0.60

Oracle

-0.60

Ĥİ

-0.60

 CLSID

-0.60

POSITIVE LOGITS

igg

1.21

imble

1.17

uggets

1.13

usra

1.09

anny

1.04

uke

1.04

ailing

1.03

ought

1.03

aught

1.02

atal

1.01

Activations Density 0.030%

words related to news or information

instances of the letter "n"

No Comments

No Known Activations