INDEX

Explanations

numerical values and measurements, particularly relating to weight and size

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

selection

-0.92

study

-0.84

repe

-0.84

reality

-0.83

sequence

-0.83

historic

-0.83

recogn

-0.83

general

-0.83

agonists

-0.82

venge

-0.81

POSITIVE LOGITS

kW

1.40

mm

1.30

kg

1.30

cm

1.29

nm

1.21

km

1.20

lbs

1.20

lb

1.18

 pounds

1.17

dB

1.16

Activations Density 0.067%

numerical values and measurements, particularly relating to weight and size

No Comments

No Known Activations