INDEX

Explanations

terms related to deceptive or exploitative practices

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs49152-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

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

area

-0.66

pole

-0.66

-0.64

canon

-0.64

areth

-0.64

shown

-0.63

REL

-0.62

cube

-0.62

ixed

-0.62

ãĤ³

-0.62

POSITIVE LOGITS

ulent

1.06

raud

1.02

gou

1.01

vertising

0.99

 extortion

0.95

ulence

0.92

 enterprises

0.91

 profits

0.88

 schemes

0.86

eering

0.84

Activations Density 0.031%

terms related to deceptive or exploitative practices

No Comments

No Known Activations