INDEX

Explanations

instances of the word "nearly" followed by a number, as well as specific detail descriptions that provide context or additional information

oai_token-act-pair · gpt-3.5-turbo

phrases that include the word "nearly" followed by a numerical qualifier

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

agate

-0.79

 Dynamics

-0.70

Reviewer

-0.68

 messenger

-0.65

plates

-0.65

":[

-0.64

Journal

-0.63

oris

-0.63

 Provision

-0.62

Plan

-0.62

POSITIVE LOGITS

 identical

0.84

 tripled

0.79

 drowned

0.76

 extinct

0.76

 doubled

0.72

 bankrupt

0.71

 unanimous

0.71

kill

0.70

sighted

0.68

 impossible

0.67

Activations Density 0.024%

instances of the word "nearly" followed by a number, as well as specific detail descriptions that provide context or additional information

phrases that include the word "nearly" followed by a numerical qualifier

No Comments

No Known Activations