INDEX

Explanations

the abbreviation "tw" followed by various sequences of characters and numbers

oai_token-act-pair · gpt-3.5-turbo

mentions of the word "twin."

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

andowski

-0.78

LAB

-0.76

ISA

-0.75

inventory

-0.72

 Marino

-0.71

 Carney

-0.71

ICA

-0.69

ARY

-0.68

idential

-0.67

HR

-0.66

POSITIVE LOGITS

tw

3.57

Tw

1.85

tw

1.58

 twitch

1.53

Tw

1.37

TW

1.36

th

1.24

TW

1.19

 wink

1.13

ele

1.12

Activations Density 0.018%

the abbreviation "tw" followed by various sequences of characters and numbers

mentions of the word "twin."

No Comments

No Known Activations