INDEX

Explanations

instances of negative political language

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

 disemb

-0.83

 hatch

-0.81

 helper

-0.74

 inactive

-0.73

 defe

-0.73

 pocket

-0.71

 carbohyd

-0.71

 transporter

-0.71

 purse

-0.70

 undet

-0.68

POSITIVE LOGITS

 Moreover

1.39

 Indeed

1.39

 Worse

1.38

 Meanwhile

1.32

 Whereas

1.31

 Likewise

1.30

Yet

1.30

 Such

1.29

 Ultimately

1.29

 Nonetheless

1.28

Activations Density 1.743%

instances of negative political language

No Comments

No Known Activations