INDEX

Explanations

references to people and their relationships

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs49152-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

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

osphere

-0.76

trop

-0.74

fecture

-0.72

abase

-0.72

availability

-0.71

flush

-0.69

orgetown

-0.69

Indust

-0.68

verage

-0.68

iti

-0.68

POSITIVE LOGITS

 selves

1.11

 brothers

1.03

 siblings

1.00

selves

0.94

 sisters

0.92

 were

0.92

 minds

0.91

duo

0.91

 reunited

0.89

 twins

0.89

Activations Density 0.262%

references to people and their relationships

No Comments

No Known Activations