INDEX

Explanations

references to job titles or positions within organizations or institutions

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

adle

-0.80

metic

-0.78

tered

-0.77

 Frie

-0.73

 tits

-0.68

roman

-0.68

 vigil

-0.68

 vanquished

-0.65

relation

-0.65

 whistle

-0.65

POSITIVE LOGITS

 Edition

0.98

 Achievement

0.94

 Fellow

0.94

 Services

0.94

 Manager

0.93

 Advisor

0.92

 Analyst

0.91

 Member

0.90

 Excellence

0.89

 Producer

0.88

Activations Density 0.015%

references to job titles or positions within organizations or institutions

No Comments

No Known Activations