INDEX

Explanations

verbs expressing action or judgement

oai_token-act-pair · gpt-3.5-turbo

phrases related to criticism and accountability

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

inger

-0.72

ilogy

-0.68

ciating

-0.68

talking

-0.65

Depending

-0.64

DragonMagazine

-0.64

Reporting

-0.63

gui

-0.62

 thanking

-0.60

orget

-0.58

POSITIVE LOGITS

 insensitive

0.77

BuyableInstoreAndOnline

0.74

 sake

0.72

 improper

0.70

ãĤ¤ãĥĪ

0.69

 insufficient

0.68

 unsu

0.65

 purposes

0.65

 illegal

0.64

 unconventional

0.64

Activations Density 0.176%

verbs expressing action or judgement

phrases related to criticism and accountability

No Comments

No Known Activations