INDEX

Explanations

references to the word "infinity."

oai_token-act-pair · gpt-3.5-turbo

suffixed terms related to concepts of identity and fidelity

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

accompan

-0.85

ozo

-0.74

umm

-0.67

abi

-0.67

awaru

-0.66

roman

-0.65

adult

-0.65

ask

-0.64

orically

-0.64

artisan

-0.64

POSITIVE LOGITS

inity

0.87

idia

0.75

 Towers

0.73

xes

0.73

 Reserved

0.72

 Hots

0.72

idad

0.72

ity

0.69

creen

0.68

 Grid

0.68

Activations Density 0.030%

references to the word "infinity."

suffixed terms related to concepts of identity and fidelity

No Comments

No Known Activations