INDEX

Explanations

instances of specific events or situations, particularly when they are negatively noteworthy

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate notable events or mentions within a context

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Negative Logits

orth

-0.67

alted

-0.64

uitive

-0.64

vantage

-0.63

INTON

-0.62

complex

-0.62

iverse

-0.60

nex

-0.60

Materials

-0.60

sonian

-0.60

POSITIVE LOGITS

 when

1.02

when

0.96

 mentioning

0.83

 during

0.82

 witnessing

0.82

during

0.80

 dismissing

0.79

 announcing

0.78

 recalling

0.78

 visits

0.77

Activations Density 0.273%

instances of specific events or situations, particularly when they are negatively noteworthy

phrases that indicate notable events or mentions within a context

No Known Activations