INDEX

Explanations

references to well-known newspapers and publications

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_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

sed

-0.87

cles

-0.76

gone

-0.75

wana

-0.74

uga

-0.72

facebook

-0.69

perm

-0.69

isable

-0.68

together

-0.66

apses

-0.64

POSITIVE LOGITS

 Editorial

1.25

 Newsp

1.19

 Magazine

1.14

 editorial

1.12

 Newspaper

1.10

 columnist

1.07

 Editors

1.02

 Dispatch

1.02

 Herald

1.00

 article

0.98

Activations Density 0.535%

references to well-known newspapers and publications

No Comments

No Known Activations