INDEX

Explanations

words related to specific names, organizations, and events

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

 compe

-0.80

PORT

-0.69

 Fargo

-0.66

 parasitic

-0.62

 Exodus

-0.62

 credit

-0.62

 Sacrament

-0.62

Ae

-0.62

 Lumpur

-0.61

 GOODMAN

-0.61

POSITIVE LOGITS

idd

1.35

umbling

1.22

iddling

1.17

umbled

1.15

ugs

1.14

ips

1.11

agg

1.11

abb

1.11

umbles

1.10

ashes

1.09

Activations Density 0.774%

words related to specific names, organizations, and events

No Comments

No Known Activations