INDEX

Explanations

the word "Com" within texts

oai_token-act-pair · gpt-3.5-turbo

mentions of the term "Com" or related entities associated with it

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 harbor

-0.79

 harbour

-0.67

 spoil

-0.60

 inund

-0.60

 recoil

-0.59

 Kubrick

-0.58

 politic

-0.58

 unlock

-0.58

 epis

-0.57

 bitters

-0.57

POSITIVE LOGITS

ptroller

1.42

rade

1.37

pleting

1.33

pletion

1.30

plet

1.27

fortable

1.27

plement

1.24

odo

1.21

anche

1.20

position

1.17

Activations Density 0.015%

the word "Com" within texts

mentions of the term "Com" or related entities associated with it

No Comments

No Known Activations