INDEX

Explanations

negations or contractions with "aren't"

oai_token-act-pair · gpt-3.5-turbo

negative contractions of the verb "are."

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

assi

-0.58

 UNHCR

-0.55

com

-0.51

amaz

-0.49

 continuous

-0.48

Ext

-0.48

 undert

-0.48

APP

-0.48

ces

-0.48

 conv

-0.48

POSITIVE LOGITS

 aren

2.84

 weren

2.57

 haven

1.78

 shouldn

1.49

isn

1.48

don

1.45

 Aren

1.42

 wouldn

1.35

ain

1.34

 hadn

1.28

Activations Density 0.018%

negations or contractions with "aren't"

negative contractions of the verb "are."

No Comments

No Known Activations