INDEX

Explanations

questions, particularly starting with the word "How"

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

room

-0.62

piece

-0.61

odder

-0.60

article

-0.59

oubted

-0.57

 hereafter

-0.57

iculture

-0.56

 Issue

-0.56

agonists

-0.56

goers

-0.56

POSITIVE LOGITS

soever

1.10

beit

0.97

ever

0.95

ells

0.91

itzer

0.90

ling

0.88

 much

0.87

ls

0.86

 exactly

0.75

much

0.75

Activations Density 1.164%

questions, particularly starting with the word "How"

No Comments

No Known Activations