INDEX

Explanations

the concept of misunderstanding or misinterpretation in text

oai_token-act-pair · gpt-3.5-turbo

references to misunderstandings and related themes

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

Pwr

-0.66

 ascend

-0.66

adder

-0.62

binding

-0.62

metics

-0.62

 explicit

-0.62

atars

-0.61

daq

-0.61

 stripes

-0.61

Action

-0.59

POSITIVE LOGITS

 misunderstanding

2.50

 misunderstand

2.14

 misunderstood

1.90

 curiosity

1.70

 misinterpret

1.59

 misunder

1.48

 misconception

1.33

 prank

1.25

 misconceptions

1.20

 diversion

1.14

Activations Density 0.057%

the concept of misunderstanding or misinterpretation in text

references to misunderstandings and related themes

No Comments

No Known Activations