INDEX

Explanations

expressions related to an entity being complete or containing all parts

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_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

KER

-0.72

chairs

-0.66

none

-0.65

claimed

-0.65

ker

-0.62

tips

-0.62

friends

-0.62

bid

-0.61

cs

-0.61

 whispers

-0.60

POSITIVE LOGITS

heartedly

1.38

hearted

1.20

 ordeal

0.97

 thing

0.93

gam

0.92

 globe

0.85

 spectrum

0.85

 darn

0.75

 ecosystem

0.74

 suite

0.74

Activations Density 0.399%

expressions related to an entity being complete or containing all parts

No Comments

No Known Activations