INDEX

Explanations

advice-related phrases and questions

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

Bom

-0.68

 Alive

-0.68

plex

-0.65

buck

-0.64

BP

-0.64

iny

-0.62

yss

-0.61

 Mellon

-0.60

atos

-0.60

yards

-0.60

POSITIVE LOGITS

 advice

1.08

 Advice

1.03

 advising

0.99

tip

0.93

 guide

0.93

 advise

0.91

 counselor

0.88

 counselors

0.88

 guides

0.85

 guidance

0.84

Activations Density 1.091%

advice-related phrases and questions

No Comments

No Known Activations