INDEX

Explanations

keywords related to communication or interaction

oai_token-act-pair · gpt-3.5-turbo

references to physical or interpersonal contact

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

udo

-0.74

Republicans

-0.67

iev

-0.66

 Built

-0.66

idy

-0.64

 roared

-0.62

èĢ

-0.62

 Throne

-0.62

 presided

-0.62

Bench

-0.61

POSITIVE LOGITS

 contact

3.92

contact

2.91

 Contact

2.75

 contacts

2.57

Contact

2.08

 contacting

2.06

 contacted

1.56

 communication

1.28

 touch

1.27

 CONT

1.21

Activations Density 0.020%

keywords related to communication or interaction

references to physical or interpersonal contact

No Comments

No Known Activations