INDEX

Explanations

lines ending with specific symbols or patterns, possibly related to formatting or metadata

oai_token-act-pair · gpt-3.5-turbo

content related to video and media representation

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 congr

-0.67

}}}

-0.65

 Kant

-0.65

 Chomsky

-0.62

 congratulations

-0.60

Ud

-0.60

 substr

-0.60

 Definitely

-0.59

 abolished

-0.59

 technically

-0.59

POSITIVE LOGITS

Philadelphia

0.76

CLE

0.74

MI

0.74

PORT

0.72

ILLE

0.72

GREEN

0.72

INGTON

0.72

ampa

0.72

CHAR

0.71

ALE

0.70

Activations Density 0.150%

lines ending with specific symbols or patterns, possibly related to formatting or metadata

content related to video and media representation

No Comments

No Known Activations