INDEX

Explanations

specific words related to a structured command or instruction, potentially in a programming context

oai_token-act-pair · gpt-3.5-turbo

mentions of conflict or struggle

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

WER

-0.68

 disapp

-0.68

gow

-0.67

jong

-0.66

Netflix

-0.65

hua

-0.65

udder

-0.65

ksh

-0.64

ndra

-0.64

reau

-0.64

POSITIVE LOGITS

Sensor

0.66

âĦ¢:

0.63

pin

0.63

 defensive

0.63

 Catch

0.62

 Powered

0.61

 Hills

0.59

 cylinder

0.57

 Wake

0.57

andowski

0.57

Activations Density 0.000%

specific words related to a structured command or instruction, potentially in a programming context

mentions of conflict or struggle

No Comments

No Known Activations