INDEX

Explanations

mentions of government actions or discussions related to government responses

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_fs768-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

768

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

 Plate

-0.71

 Candle

-0.68

hots

-0.68

 Heaven

-0.67

Pie

-0.65

 Angle

-0.65

 Celtic

-0.63

 Width

-0.63

Sic

-0.61

ocalyptic

-0.61

POSITIVE LOGITS

 agencies

1.18

 bureaucracy

1.17

 officials

1.14

 bureaucrats

1.07

 sanctioned

1.02

 coffers

0.96

 shutdown

0.95

 Accountability

0.94

 agency

0.92

 mandated

0.91

Activations Density 3.235%

mentions of government actions or discussions related to government responses

No Comments

No Known Activations