INDEX

Explanations

television network names and abbreviations

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

taboola

-0.87

 thous

-0.59

 showc

-0.59

 puff

-0.59

give

-0.58

otrop

-0.58

 trave

-0.58

 surpr

-0.58

 Allaah

-0.57

 soph

-0.57

POSITIVE LOGITS

KA

1.10

KT

1.04

OPS

0.98

AR

0.97

WT

0.96

HY

0.95

OT

0.95

NT

0.94

KI

0.93

ATA

0.93

Activations Density 1.657%

television network names and abbreviations

No Comments

No Known Activations