INDEX

Explanations

phrases related to past actions or events

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

bie

-0.62

PI

-0.62

âĻ¦

-0.61

bery

-0.60

hack

-0.60

owe

-0.57

hammer

-0.57

 Voters

-0.57

 trope

-0.56

etting

-0.56

POSITIVE LOGITS

 been

1.35

 undergone

1.15

 begun

1.15

 gone

1.10

 gotten

1.09

iths

1.07

been

1.02

 previously

0.98

 flown

0.94

 taken

0.91

Activations Density 0.651%

phrases related to past actions or events

No Comments

No Known Activations