INDEX

Explanations

mentions of specific publications or news outlets

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

wana

-0.76

ABE

-0.75

urtles

-0.69

ament

-0.68

etime

-0.66

achev

-0.66

zsche

-0.65

ayer

-0.65

ements

-0.65

gradient

-0.64

POSITIVE LOGITS

 Chronicle

1.05

 Herald

0.92

naire

0.81

Magazine

0.71

 Books

0.70

itect

0.69

 Newsp

0.69

mare

0.68

Phoenix

0.66

Pages

0.65

Activations Density 0.019%

mentions of specific publications or news outlets

No Comments

No Known Activations