INDEX

Explanations

historical events and accomplishments

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

 currently

-0.73

 WATCH

-0.73

zbollah

-0.71

claimer

-0.71

 varies

-0.70

 depends

-0.69

currently

-0.65

scribe

-0.65

aka

-0.65

Latest

-0.65

POSITIVE LOGITS

 consisted

1.02

didn

0.99

 hadn

0.94

 proceeded

0.93

 knew

0.89

Had

0.88

Was

0.86

 seemed

0.85

 mattered

0.85

 didn

0.84

Activations Density 11.750%

historical events and accomplishments

No Comments

No Known Activations