INDEX

Explanations

words related to specific names and entities

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

ovember

-0.76

ONEY

-0.75

ATIONAL

-0.75

beard

-0.73

strap

-0.73

stakes

-0.69

DAY

-0.68

acion

-0.66

FORE

-0.64

astern

-0.63

POSITIVE LOGITS

ught

1.01

eger

0.92

reys

0.79

ionage

0.78

inished

0.77

isson

0.75

illard

0.71

plets

0.71

ibilities

0.71

Ń·

0.69

Activations Density 0.037%

words related to specific names and entities

No Comments

No Known Activations