INDEX

Explanations

phrases related to political or legal actions

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

 srfAttach

-0.77

Lag

-0.73

 Relief

-0.73

PsyNetMessage

-0.69

Wid

-0.66

 Times

-0.64

 Taxes

-0.64

 Mobility

-0.64

 Elements

-0.63

 Wick

-0.63

POSITIVE LOGITS

 replaced

1.10

 able

1.08

 evaluated

1.07

 investigated

1.03

 judged

1.03

 transported

1.01

 phased

1.00

 assessed

1.00

 treated

0.99

 tested

0.99

Activations Density 0.482%

phrases related to political or legal actions

No Comments

No Known Activations