INDEX

Explanations

names of characters or subjects related to entertainment and media

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 convenient

-0.57

pects

-0.56

 sufficiently

-0.56

 nowadays

-0.55

 rewarding

-0.55

 mandate

-0.53

 acutely

-0.51

.--

-0.51

 worthwhile

-0.51

 lightly

-0.51

POSITIVE LOGITS

etc

0.83

ife

0.60

etc

0.59

owa

0.59

rup

0.58

rote

0.57

idth

0.57

bra

0.56

iba

0.55

eka

0.53

Activations Density 0.358%

names of characters or subjects related to entertainment and media

No Comments

No Known Activations