INDEX

Explanations

names of people and positions

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_fs768-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

768

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

Decre

-0.66

 progress

-0.64

 necessities

-0.62

 assumptions

-0.61

 correctness

-0.60

 trunc

-0.60

Characters

-0.60

 stereotypes

-0.60

 versa

-0.60

 ital

-0.60

POSITIVE LOGITS

Jr

1.27

Sr

1.13

PhD

1.04

 founder

0.99

 grandson

0.97

 chairman

0.95

QC

0.95

founder

0.94

 Founder

0.92

 director

0.91

Activations Density 3.951%

names of people and positions

No Comments

No Known Activations