INDEX

Explanations

words related to politics and news reporting

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_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

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

istry

-0.72

etsy

-0.68

quez

-0.68

lda

-0.68

urion

-0.66

wikipedia

-0.65

cles

-0.64

 Fraz

-0.64

potion

-0.62

 Luigi

-0.61

POSITIVE LOGITS

 broadcasts

1.38

 aired

1.22

 airing

1.19

 broadcaster

1.17

 broadcast

1.17

 television

1.17

levision

1.14

 viewership

1.13

 airs

1.11

 Broadcast

1.11

Activations Density 2.007%

words related to politics and news reporting

No Comments

No Known Activations