INDEX

Explanations

instances of bracketed content or annotations

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

Ļ½

-0.73

©¶æ¥µ

-0.71

elsh

-0.69

-+-+

-0.67

ãĥ¼ãĥĨ

-0.66

ĪĴ

-0.65

ENN

-0.65

ĻĤ

-0.64

Ń·

-0.64

 earthqu

-0.62

POSITIVE LOGITS

."[

0.73

selves

0.70

},"

0.69

".[

0.68

,"

0.67

alone

0.66

cape

0.63

thumbnails

0.63

.")

0.61

],

0.59

Activations Density 0.032%

instances of bracketed content or annotations

No Comments

No Known Activations