INDEX

Explanations

times or specific periods

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

 Amen

-0.77

 interven

-0.70

aukee

-0.64

 wills

-0.64

 sexes

-0.61

BOOK

-0.61

ieth

-0.60

ifice

-0.60

 regenerate

-0.60

 Constitution

-0.59

POSITIVE LOGITS

 understandably

0.92

 rightfully

0.91

 unsurprisingly

0.86

 predictably

0.84

 rightly

0.79

Yesterday

0.75

 Turns

0.75

Apparently

0.74

Unfortunately

0.74

 Unfortunately

0.73

Activations Density 0.712%

times or specific periods

No Comments

No Known Activations