INDEX

Explanations

the word "all" with a high level of activation

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

IDS

-0.66

SHIP

-0.63

bledon

-0.62

 Kamp

-0.62

sofar

-0.61

 Caption

-0.61

KH

-0.60

 Provision

-0.59

plin

-0.58

abwe

-0.57

POSITIVE LOGITS

igator

1.29

ocating

1.23

usion

1.13

igators

1.12

usions

1.05

uring

1.04

usive

1.00

iance

1.00

iances

0.99

edged

0.98

Activations Density 0.032%

the word "all" with a high level of activation

No Comments

No Known Activations