INDEX

Explanations

numerical information, especially related to dates or quantities

oai_token-act-pair · gpt-3.5-turbo

end-of-sequence markers in the text

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

 manif

-0.82

 synerg

-0.79

 allied

-0.76

 manifold

-0.76

 concentrating

-0.75

 prolifer

-0.75

 mounted

-0.75

 endeav

-0.75

 branches

-0.74

 harness

-0.74

POSITIVE LOGITS

ï¸ı

1.38

âĢ

1.20

 âĢº

1.15

âĸ¬âĸ¬

1.14

âĸ

1.13

âĶĢâĶĢâĶĢâĶĢâĶĢâĶĢâĶĢâĶĢ

1.10

âĹ

1.09

à¨

1.09

 à¤

1.05

Û

1.04

Activations Density 0.090%

numerical information, especially related to dates or quantities

end-of-sequence markers in the text

No Comments

No Known Activations

numerical information, especially related to dates or quantities

end-of-sequence markers in the text

No Comments

No Known Activations