INDEX

Explanations

names or terms related to business or politics

oai_token-act-pair · gpt-3.5-turbo 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

 withholding

-0.74

 almonds

-0.66

 Highlander

-0.64

 preferential

-0.63

 behold

-0.63

¿½

-0.62

 uranium

-0.61

 school

-0.61

 shroud

-0.60

 permanent

-0.59

POSITIVE LOGITS

ultane

1.59

psons

1.39

iliar

1.38

ulators

1.35

ulations

1.35

pson

1.30

ply

1.26

pler

1.24

ulator

1.22

ulation

1.22

Activations Density 0.024%

names or terms related to business or politics

No Comments

No Known Activations