INDEX

Explanations

dates in the format "dd/mm/yyyy."

oai_token-act-pair · gpt-3.5-turbo

specific years and numerical data

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

helps

-0.84

conservancy

-0.82

tackle

-0.80

pressure

-0.80

omething

-0.78

argon

-0.78

Compan

-0.77

BUS

-0.76

Pope

-0.74

Planet

-0.73

POSITIVE LOGITS

TBA

1.05

1.03

TBD

1.00

0.97

0.95

0.91

 ãĤ

0.91

 âĪ

0.90

Activations Density 0.127%

dates in the format "dd/mm/yyyy."

specific years and numerical data

No Comments

No Known Activations