INDEX

Explanations

past tense verbs related to actions, events, or incidents

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

Zone

-0.55

 Stability

-0.54

 formation

-0.53

 stripes

-0.53

 Clicker

-0.52

 fortunes

-0.52

talk

-0.51

 reinforcement

-0.51

ierre

-0.51

 Confeder

-0.50

POSITIVE LOGITS

by

1.45

 aback

1.12

BY

1.09

by

0.99

pez

0.87

uled

0.87

By

0.85

 separately

0.83

igated

0.81

By

0.80

Activations Density 2.974%

past tense verbs related to actions, events, or incidents

No Comments

No Known Activations