INDEX

Explanations

mentions of official or formal names, titles, and acronyms

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_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

-0.89

bat

-0.64

hold

-0.61

beh

-0.60

otrop

-0.59

lying

-0.59

 thous

-0.59

spe

-0.58

Ã§

-0.58

 clutter

-0.57

POSITIVE LOGITS

KA

1.16

OPS

1.11

AR

1.08

KT

1.07

KI

1.06

WT

1.05

OT

1.05

TO

1.03

HY

1.02

IA

1.01

Activations Density 0.848%

mentions of official or formal names, titles, and acronyms

No Comments

No Known Activations