INDEX

Explanations

surnames, potentially from news articles or academic works

oai_token-act-pair · gpt-3.5-turbo

names of people and places

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.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

 referen

-0.62

$$$$

-0.52

 dime

-0.51

 thous

-0.51

 ancest

-0.49

 redistributed

-0.49

 sovere

-0.47

 notch

-0.45

 predec

-0.45

 taxp

-0.44

POSITIVE LOGITS

ank

0.61

hus

0.58

ats

0.57

ias

0.56

ius

0.56

oop

0.56

bek

0.56

gal

0.56

oli

0.55

ema

0.54

Activations Density 1.068%

surnames, potentially from news articles or academic works

names of people and places

No Comments

No Known Activations