INDEX

Explanations

statements related to political figures and their actions or behaviors

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

 slee

-0.63

board

-0.61

ilated

-0.60

 Travels

-0.59

apr

-0.56

 rush

-0.56

Offline

-0.55

 awoke

-0.55

 scrimmage

-0.53

aw

-0.53

POSITIVE LOGITS

nor

1.67

Nor

1.38

 Instead

1.25

Instead

1.25

 Rather

1.24

nor

1.23

 merely

1.19

Rather

1.17

Nor

1.16

Nevertheless

1.10

Activations Density 5.267%

statements related to political figures and their actions or behaviors

No Comments

No Known Activations