INDEX

Explanations

names related to political figures or entities

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ATTLE

-0.78

 COVER

-0.75

 Uncharted

-0.69

 cryst

-0.68

LEASE

-0.68

acebook

-0.65

 Colossus

-0.65

 unforeseen

-0.62

 readable

-0.62

 redundancy

-0.61

POSITIVE LOGITS

hai

1.29

ji

1.14

gui

1.13

azi

1.13

uddin

1.12

uni

1.10

gha

1.08

 Hasan

1.07

hi

1.07

si

1.06

Activations Density 2.249%

names related to political figures or entities

No Comments

No Known Activations