INDEX

Explanations

specific labels or classifications with numbers and letters mixed in

oai_token-act-pair · gpt-3.5-turbo

classifications or categories of items

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

enhagen

-0.60

 Notting

-0.57

annels

-0.55

ipedia

-0.50

trop

-0.50

âĸĪâĸĪâĸĪâĸĪ

-0.50

leys

-0.50

interrupted

-0.49

abad

-0.48

âĸĪâĸĪ

-0.48

POSITIVE LOGITS

inctions

0.55

ularity

0.52

 guiActiveUnfocused

0.52

 honors

0.50

ciation

0.49

holder

0.49

holders

0.49

 Marketable

0.48

 Individuals

0.48

 Males

0.48

Activations Density 1.329%

specific labels or classifications with numbers and letters mixed in

classifications or categories of items

No Comments

No Known Activations