INDEX

Explanations

references to soccer/football related terms and player transactions

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_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

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

 Navajo

-1.09

colored

-1.07

GOP

-1.05

 Cherokee

-0.99

FBI

-0.99

 Whitman

-0.97

 Hawai

-0.96

 nonprofits

-0.94

DOD

-0.94

DOE

-0.94

POSITIVE LOGITS

 Â£

1.70

 footballer

1.64

 realise

1.60

 centre

1.56

 Newcastle

1.54

 enqu

1.54

 recognise

1.54

 Liverpool

1.53

 analyse

1.52

 organising

1.51

Activations Density 7.518%

references to soccer/football related terms and player transactions

No Comments

No Known Activations