INDEX

Explanations

names of political figures and locations

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_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

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

:(

-0.69

esthetic

-0.66

DIT

-0.64

fiction

-0.63

Redditor

-0.63

FIG

-0.62

pleasant

-0.62

":"/

-0.62

fff

-0.61

advertisement

-0.61

POSITIVE LOGITS

 others

1.08

 assorted

0.98

Kw

0.90

Zac

0.85

Tel

0.84

 Trent

0.81

etc

0.80

 Trey

0.80

etc

0.78

 possibly

0.77

Activations Density 0.121%

names of political figures and locations

No Comments

No Known Activations