INDEX

Explanations

names followed by a verb or action, or simply a call for action

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 istg

-0.70

inth

-0.70

bilt

-0.65

————

-0.64

Ģ

-0.63

bourne

-0.63

olitics

-0.62

±

-0.61

isphere

-0.59

 guiName

-0.59

POSITIVE LOGITS

igraph

1.13

backs

1.09

0.97

 attention

0.96

 bluff

0.88

 forth

0.84

call

0.83

aghan

0.82

oused

0.82

 bullshit

0.80

Activations Density 1.260%

names followed by a verb or action, or simply a call for action

No Comments

No Known Activations