INDEX

Explanations

words related to media, press, public, internet, laws, amendments, and legal institutions

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

¶æ

-0.77

ĪĴ

-0.72

edIn

-0.71

ãĤ½

-0.69

ļé

-0.68

Ô

-0.66

sbm

-0.66

inventoryQuantity

-0.66

ĨĴ

-0.64

ãĤ¨ãĥ«

-0.63

POSITIVE LOGITS

 itself

0.99

osphere

0.92

 hierarchy

0.82

liest

0.82

 Museum

0.79

 Establishment

0.77

mma

0.77

 Gazette

0.76

era

0.76

 logo

0.76

Activations Density 3.495%

words related to media, press, public, internet, laws, amendments, and legal institutions

No Comments

No Known Activations