INDEX

Explanations

phrases involving both sides agreeing, disagreeing, or being in conflict

oai_token-act-pair · gpt-3.5-turbo

references to opposing parties or groups involved in a negotiation or conflict

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Negative Logits

cles

-0.64

heon

-0.64

 Berry

-0.63

 Countdown

-0.63

 Miko

-0.62

 dehyd

-0.62

UCH

-0.60

Lup

-0.60

 Patron

-0.59

uate

-0.59

POSITIVE LOGITS

side

0.82

pace

0.80

 sides

0.75

 thereof

0.74

 concede

0.73

gaard

0.71

itution

0.71

thro

0.71

ãĥķãĤ©

0.71

 handshake

0.70

Activations Density 0.028%

phrases involving both sides agreeing, disagreeing, or being in conflict

references to opposing parties or groups involved in a negotiation or conflict

No Known Activations