INDEX

Explanations

indications of potential problems or anomalies

oai_token-act-pair · gpt-3.5-turbo

phrases indicating a problem or something amiss

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

Not in Any Lists

No Comments

Negative Logits

 praises

-0.75

icio

-0.64

 reviews

-0.63

 fame

-0.63

vale

-0.62

 Rican

-0.62

 Documents

-0.60

ulia

-0.60

 predecessors

-0.59

 cites

-0.59

POSITIVE LOGITS

 wrong

1.28

wrong

1.14

 terribly

1.02

 horribly

1.00

 bothering

0.99

 happening

0.95

 rotten

0.93

 Wrong

0.90

missing

0.90

 missing

0.88

Activations Density 0.104%

indications of potential problems or anomalies

phrases indicating a problem or something amiss

No Comments

No Known Activations