INDEX

Explanations

phrases related to news reporting or journalism

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_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

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

wives

-0.65

emort

-0.64

naissance

-0.63

tre

-0.63

lection

-0.61

nerg

-0.61

thing

-0.61

ãĥŃ

-0.59

ordial

-0.59

clamation

-0.58

POSITIVE LOGITS

ometimes

0.69

 petertodd

0.69

olate

0.67

 favorably

0.66

 quotes

0.66

 citing

0.64

metics

0.63

 omin

0.63

 anecdotes

0.63

 "...

0.62

Activations Density 0.205%

phrases related to news reporting or journalism

No Comments

No Known Activations