INDEX

Explanations

terms related to data filtering and manipulation

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

eanor

-0.79

ciating

-0.76

arrass

-0.70

ILY

-0.68

rious

-0.68

erald

-0.67

ocamp

-0.66

verty

-0.66

ynes

-0.66

ston

-0.66

POSITIVE LOGITS

 filters

0.90

 filter

0.88

 filtering

0.87

filter

0.81

operator

0.80

 Filter

0.80

 cutoff

0.79

 sweeps

0.78

 filtered

0.69

 curtains

0.68

Activations Density 0.106%

terms related to data filtering and manipulation

No Comments

No Known Activations