INDEX

Explanations

acronyms and abbreviations related to organizations and entities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

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

orer

-0.71

fusc

-0.68

phasis

-0.62

ubis

-0.62

idae

-0.61

Redditor

-0.60

 istg

-0.60

roman

-0.59

inator

-0.59

owicz

-0.58

POSITIVE LOGITS

VILLE

1.46

 CITY

1.25

VIEW

1.15

LAND

1.14

FIELD

1.13

 COUNTY

1.11

BUR

1.11

INGTON

1.10

STON

1.10

WOOD

1.08

Activations Density 0.113%

acronyms and abbreviations related to organizations and entities

No Comments

No Known Activations