INDEX

Explanations

words related to being ambushed or being on guard against danger

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

leaf

-0.68

PIN

-0.67

 Dakota

-0.67

 twins

-0.66

terday

-0.65

 Parenthood

-0.64

Nun

-0.62

 Doctors

-0.62

 Ceres

-0.61

 AMERICA

-0.61

POSITIVE LOGITS

olic

1.01

ivalent

0.99

iance

0.98

ience

0.97

isexual

0.96

ushed

0.95

rill

0.95

etr

0.94

otiation

0.94

olin

0.94

Activations Density 0.025%

words related to being ambushed or being on guard against danger

No Comments

No Known Activations