INDEX

Explanations

mentions of geographical locations, particularly military-related

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

PsyNetMessage

-0.91

icular

-0.84

BIL

-0.79

OPLE

-0.72

ATIONS

-0.71

atically

-0.70

ript

-0.70

rative

-0.68

ICLE

-0.68

oldemort

-0.67

POSITIVE LOGITS

berry

1.15

smith

1.12

hawk

1.11

ened

1.05

adder

1.05

bird

1.05

feet

1.03

jack

1.02

beard

0.99

 Lives

0.98

Activations Density 0.025%

mentions of geographical locations, particularly military-related

No Comments

No Known Activations