INDEX

Explanations

references to political inquiries or investigations

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

 Cosponsors

-0.21

Ó

-0.20

 horm

-0.20

 sophistication

-0.20

 advoc

-0.20

 srfAttach

-0.20

taboola

-0.20

attribute

-0.20

Engineers

-0.19

POSITIVE LOGITS

hower

0.21

 anymore

0.20

ipop

0.20

puter

0.20

 actual

0.20

nant

0.20

 consecutive

0.20

ardless

0.19

:(

0.19

 straight

0.19

Activations Density 19.046%

references to political inquiries or investigations

No Comments

No Known Activations