INDEX

Explanations

phrases indicating support and presence

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_fs12288-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_12288

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

12,288

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

 reliance

-0.69

 subp

-0.65

 favour

-0.65

GF

-0.64

 lymph

-0.64

 cardinal

-0.63

 prostate

-0.59

 divisions

-0.59

 affiliates

-0.58

 Flavoring

-0.58

POSITIVE LOGITS

abouts

1.00

enced

0.80

tical

0.78

 waiting

0.78

 somewhere

0.78

oing

0.76

itored

0.76

 Somewhere

0.74

ivating

0.73

enta

0.73

Activations Density 0.196%

phrases indicating support and presence

No Comments

No Known Activations