INDEX

Explanations

references to a specific video game called "Reflex" and related gameplay modes and features

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

 Peninsula

-0.78

 thirds

-0.72

butt

-0.72

ÂłÂłÂłÂł

-0.63

 nerv

-0.61

 Wilde

-0.59

 darn

-0.59

 acceptance

-0.58

 Klux

-0.58

 custody

-0.58

POSITIVE LOGITS

ively

1.91

ions

1.73

es

1.57

ivity

1.55

iveness

1.54

ors

1.50

ive

1.42

ives

1.37

ional

1.34

ures

1.25

Activations Density 1.883%

references to a specific video game called "Reflex" and related gameplay modes and features

No Comments

No Known Activations