INDEX

Explanations

non-alphabet characters and commas

oai_token-act-pair · gpt-3.5-turbo

important statements or assertions related to technical or conceptual discussions

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 amnesty

-0.70

 whistle

-0.68

 hust

-0.63

ÃŃ

-0.59

 Aval

-0.59

 sacked

-0.59

 bomb

-0.58

 Reincarnated

-0.57

 footing

-0.56

eteen

-0.56

POSITIVE LOGITS

Therefore

1.07

Example

1.06

Furthermore

1.02

Thus

0.99

Likewise

0.99

Moreover

0.98

Similarly

0.96

Lastly

0.95

Additionally

0.94

Different

0.93

Activations Density 0.621%

non-alphabet characters and commas

important statements or assertions related to technical or conceptual discussions

No Comments

No Known Activations