INDEX

Explanations

Twitter handles along with related text information

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_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

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

 specificity

-0.81

 favors

-0.69

 compress

-0.68

 wards

-0.68

 Norwich

-0.67

 Takeru

-0.67

 orche

-0.67

 Romanian

-0.66

 Sloven

-0.66

 tape

-0.66

POSITIVE LOGITS

realDonaldTrump

1.30

thereal

1.17

Real

1.08

username

1.03

Coach

1.00

#$

0.99

Official

0.97

nat

0.97

meg

0.96

gmail

0.96

Activations Density 0.837%

Twitter handles along with related text information

No Comments

No Known Activations