INDEX

Explanations

phrases related to carrying out 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_fs6144-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_6144

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

6,144

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

ellen

-0.75

chrom

-0.74

reddits

-0.70

 Seym

-0.68

 Skydragon

-0.65

Pie

-0.64

Hill

-0.62

IOR

-0.62

Trust

-0.61

ederation

-0.61

POSITIVE LOGITS

 loads

0.84

forward

0.83

 forward

0.83

weight

0.81

 onward

0.80

 weight

0.77

 carrier

0.76

 luggage

0.74

over

0.73

 heavier

0.73

Activations Density 0.229%

phrases related to carrying out actions

No Comments

No Known Activations